AndrewKof commited on
Commit
5b11294
·
1 Parent(s): c9cc0ca

🚀 Update UI with LFS for images and models

Browse files
.DS_Store CHANGED
Binary files a/.DS_Store and b/.DS_Store differ
 
.gitattributes CHANGED
@@ -34,3 +34,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  *.png filter=lfs diff=lfs merge=lfs -text
 
 
 
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  *.png filter=lfs diff=lfs merge=lfs -text
37
+ *.jpg filter=lfs diff=lfs merge=lfs -text
38
+ *.jpeg filter=lfs diff=lfs merge=lfs -text
app/.DS_Store CHANGED
Binary files a/app/.DS_Store and b/app/.DS_Store differ
 
app/Inference_YoLOv11.py ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os
2
+ import numpy as np
3
+ from ultralytics import YOLO
4
+ from PIL import Image
5
+ import torch
6
+ import yaml
7
+
8
+ # -----------------------------
9
+ # CONFIGURATION
10
+ # -----------------------------
11
+ MODEL_PATH = "best.pt"
12
+ DATA_YAML = "data.yaml"
13
+ IMG_SIZE = 512
14
+ IMG_PATH = "chroms.jpeg"
15
+ USE_GPU = True
16
+
17
+ # -----------------------------
18
+ # GPU/CPU
19
+ # -----------------------------
20
+ if USE_GPU and torch.cuda.is_available():
21
+ torch.cuda.set_per_process_memory_fraction(0.5, device=0)
22
+ else:
23
+ torch.device('cpu')
24
+
25
+ # -----------------------------
26
+ # Load class names
27
+ # -----------------------------
28
+ with open(DATA_YAML, "r") as f:
29
+ data = yaml.safe_load(f)
30
+
31
+ class_names = data["names"] # alphabetical list
32
+ num_classes = len(class_names)
33
+
34
+ # -----------------------------
35
+ # Load YOLO Model
36
+ # -----------------------------
37
+ model = YOLO(MODEL_PATH)
38
+ print(f"Loaded model: {MODEL_PATH}")
39
+
40
+ # -----------------------------
41
+ # Build numeric folder → class index mapping
42
+ # -----------------------------
43
+ yolo_names = model.names # dictionary from YOLO output
44
+
45
+ # Convert yolo class order into integers matching species indices
46
+ mapping = {yolo_idx: int(name_str) for yolo_idx, name_str in yolo_names.items()}
47
+
48
+ # -----------------------------
49
+ # Load & preprocess image
50
+ # -----------------------------
51
+ img = Image.open(IMG_PATH).convert('RGB')
52
+ img = img.resize((IMG_SIZE, IMG_SIZE))
53
+
54
+ # -----------------------------
55
+ # Inference
56
+ # -----------------------------
57
+ results = model.predict(source=img, imgsz=IMG_SIZE, verbose=False)
58
+ probs = results[0].probs.data.cpu().numpy()
59
+ # -----------------------------
60
+ # TOP-5 OUTPUT WITH REAL NAMES
61
+ # -----------------------------
62
+ top5_yolo_idx = np.argsort(probs)[-5:][::-1]
63
+ top5_probs = probs[top5_yolo_idx] * 100
64
+
65
+ print("\n🎣 Top-5 Predictions:")
66
+ for rank, yolo_idx in enumerate(top5_yolo_idx, 1):
67
+ species_idx = mapping[yolo_idx] # map YOLO index to species index
68
+ species_name = class_names[species_idx]
69
+ print(f"{rank}. {species_name} — {top5_probs[rank-1]:.2f}%")
70
+
app/__pycache__/inference.cpython-310.pyc CHANGED
Binary files a/app/__pycache__/inference.cpython-310.pyc and b/app/__pycache__/inference.cpython-310.pyc differ
 
app/__pycache__/inference_yolo.cpython-310.pyc ADDED
Binary file (4.97 kB). View file
 
app/__pycache__/main.cpython-310.pyc CHANGED
Binary files a/app/__pycache__/main.cpython-310.pyc and b/app/__pycache__/main.cpython-310.pyc differ
 
app/__pycache__/model.cpython-310.pyc CHANGED
Binary files a/app/__pycache__/model.cpython-310.pyc and b/app/__pycache__/model.cpython-310.pyc differ
 
app/atherina.jpg ADDED

Git LFS Details

  • SHA256: ed1122809052d71eac6c187a52d06c506a3deaf7c1306569b80136a6bdff90ca
  • Pointer size: 130 Bytes
  • Size of remote file: 87.2 kB
app/best.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac6eb60c3fee4117e630b2141ddc86ede8521adba528890b2faa82754de2e335
3
+ size 57260593
app/best_yolov11n_model_r2a4v2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bfb5869b5546507b2ec1f615d960aa6eaffc3eb227de461869c27ab4d04c3564
3
+ size 6863792
app/chroms.jpeg ADDED

Git LFS Details

  • SHA256: e37bc6b72f2e95d541da4be682739660a2bb6e9e44b04673c745a0a4a530369c
  • Pointer size: 130 Bytes
  • Size of remote file: 23.5 kB
app/data.yaml ADDED
@@ -0,0 +1,107 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ path: /home/andrew/Desktop/eleni_folder/YAML_DATA
2
+ train: train
3
+ val: val
4
+ test: test
5
+ nc: 101
6
+ names:
7
+ - Aidablennius sphynx
8
+ - Apogon imberbis
9
+ - Arnoglossus laterna
10
+ - Atherina boyeri
11
+ - Atherina hepsetus
12
+ - Boops boops
13
+ - Bothus podas
14
+ - Callionymus pusillus
15
+ - Callionymus risso
16
+ - Caranx crysos
17
+ - Chelidonichthys lucerna
18
+ - Chromis chromis
19
+ - Coris julis
20
+ - Dactylopterus volitans
21
+ - Dasyatis pastinaca
22
+ - Diplodus annularis
23
+ - Diplodus puntazzo
24
+ - Diplodus sargus
25
+ - Diplodus vulgaris
26
+ - Echiichthys vipera
27
+ - Epinephelus costae
28
+ - Epinephelus marginatus
29
+ - Fistularia commersonii
30
+ - Gobius auratus
31
+ - Gobius bucchichi
32
+ - Gobius cobitis
33
+ - Gobius cruentatus
34
+ - Gobius geniporus
35
+ - Gobius incognitus
36
+ - Gobius niger
37
+ - Gobius paganellus
38
+ - Gobius vittatus
39
+ - Hippocampus guttulatus
40
+ - Hippocampus hippocampus
41
+ - Labrus bergylta
42
+ - Labrus merula
43
+ - Labrus mixtus
44
+ - Labrus viridis
45
+ - Lagocephalus sceleratus
46
+ - Lithognathus mormyrus
47
+ - Microlipophrys nigriceps
48
+ - Mugil auratus
49
+ - Mugil cephalus
50
+ - Mullus barbatus
51
+ - Mullus surmuletus
52
+ - Muraena helena
53
+ - Oblada melanura
54
+ - Pagrus pagrus
55
+ - Parablennius gattorugine
56
+ - Parablennius incognitus
57
+ - Parablennius pilicornis
58
+ - Parablennius rouxi
59
+ - Parablennius sanguinolentus
60
+ - Parablennius tentacularis
61
+ - Parablennius zvonimiri
62
+ - Parupeneus forsskali
63
+ - Phycis phycis
64
+ - Plotosus lineatus
65
+ - Pomatoschistus marmoratus
66
+ - Pterois miles
67
+ - Salaria pavo
68
+ - Sargocentron rubrum
69
+ - Sarpa salpa
70
+ - Sciaena umbra
71
+ - Scorpaena maderensis
72
+ - Scorpaena notata
73
+ - Scorpaena scrofa
74
+ - Serranus cabrilla
75
+ - Serranus hepatus
76
+ - Serranus scriba
77
+ - Siganus luridus
78
+ - Siganus rivulatus
79
+ - Solea solea
80
+ - Sparisoma cretense
81
+ - Sparus aurata
82
+ - Spicara maena
83
+ - Spicara smaris
84
+ - Spondyliosoma cantharus
85
+ - Stephanolepis diaspros
86
+ - Symphodus cinereus
87
+ - Symphodus mediterraneus
88
+ - Symphodus melanocercus
89
+ - Symphodus ocellatus
90
+ - Symphodus roissali
91
+ - Symphodus rostratus
92
+ - Symphodus tinca
93
+ - Syngnathus abaster
94
+ - Syngnathus tenuirostris
95
+ - Syngnathus typhle
96
+ - Thalassoma pavo
97
+ - Thorogobius ephippiatus
98
+ - Torpedo marmorata
99
+ - Torquigener flavimaculosus
100
+ - Trachinus araneus
101
+ - Trachinus draco
102
+ - Trachinus radiatus
103
+ - Tripterygion delaisi
104
+ - Tripterygion melanurus
105
+ - Tripterygion tripteronotus
106
+ - Uranoscopus scaber
107
+ - Xyrichtys novacula
app/inference.py CHANGED
@@ -1,99 +1,3 @@
1
- # # app/inference.py
2
- # import os
3
- # import io
4
- # import json
5
- # import torch
6
- # import torch.nn.functional as F
7
- # from PIL import Image
8
- # from transformers import AutoImageProcessor, AutoModelForImageClassification
9
-
10
- # # ─────────────────────────────────────────────
11
- # # CONFIG
12
- # # ─────────────────────────────────────────────
13
- # # Hugging Face repo for the trained checkpoint
14
- # MODEL_REPO = "Arew99/dinov2-costum"
15
-
16
- # # optional: path to local label map (bundled in your repo)
17
- # ID2NAME_PATH = os.path.join(os.path.dirname(__file__), "id2name.json")
18
-
19
- # # detect device automatically
20
- # DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
21
- # print(f"🧠 Using device: {DEVICE}")
22
-
23
- # # global cache (so model loads only once)
24
- # _model = None
25
- # _processor = None
26
- # _id2name = None
27
-
28
-
29
- # # ─────────────────────────────────────────────
30
- # # HELPER: load label map
31
- # # ─────────────────────────────────────────────
32
- # def _load_id2name():
33
- # if os.path.exists(ID2NAME_PATH):
34
- # with open(ID2NAME_PATH, "r") as f:
35
- # data = json.load(f)
36
- # # ensure integer keys
37
- # return {int(k): v for k, v in data.items()}
38
- # print("⚠️ id2name.json not found — using placeholder labels.")
39
- # return {i: f"Class {i}" for i in range(101)} # fallback
40
-
41
-
42
- # # ─────────────────────────────────────────────
43
- # # INIT: load model & processor once
44
- # # ─────────────────────────────────────────────
45
- # def load_classification_model():
46
- # global _model, _processor, _id2name
47
-
48
- # if _model is not None:
49
- # return _model, _processor, _id2name
50
-
51
- # print(f"🔁 Loading model from Hugging Face repo: {MODEL_REPO}")
52
- # _processor = AutoImageProcessor.from_pretrained(MODEL_REPO)
53
- # _model = AutoModelForImageClassification.from_pretrained(
54
- # MODEL_REPO,
55
- # ignore_mismatched_sizes=True,
56
- # ).to(DEVICE)
57
- # _model.eval()
58
- # _id2name = _load_id2name()
59
-
60
- # print(f"✅ Model loaded and ready on {DEVICE}")
61
- # return _model, _processor, _id2name
62
-
63
-
64
- # # ─────────────────────────────────────────────
65
- # # INFERENCE: classify raw image bytes
66
- # # ─────────────────────────────────────────────
67
- # def classify_bytes(image_bytes: bytes):
68
- # model, processor, id2name = load_classification_model()
69
-
70
- # image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
71
- # inputs = processor(images=image, return_tensors="pt").to(DEVICE)
72
-
73
- # with torch.no_grad():
74
- # outputs = model(**inputs)
75
- # probs = F.softmax(outputs.logits, dim=-1)
76
-
77
- # topk = torch.topk(probs, k=5)
78
- # indices = topk.indices[0].tolist()
79
- # values = topk.values[0].tolist()
80
-
81
- # results = []
82
- # for rank, (idx, prob) in enumerate(zip(indices, values), 1):
83
- # label = id2name.get(int(idx), f"Class {idx}")
84
- # results.append({
85
- # "rank": rank,
86
- # "id": int(idx),
87
- # "label": label,
88
- # "score": float(prob),
89
- # })
90
-
91
- # # concise summary for API
92
- # return {
93
- # "top1": results[0],
94
- # "top5": results,
95
- # }
96
-
97
  # app/inference.py
98
  import os
99
  import io
@@ -204,7 +108,7 @@ def classify_bytes(image_bytes: bytes):
204
  edgecolor="none", height=0.6)
205
  ax[1].set_xlim(0, 100)
206
  ax[1].set_xlabel("Probability (%)", fontsize=11)
207
- ax[1].set_title("Top-3 Predicted Species", fontsize=12, weight="bold")
208
 
209
  for bar, prob in zip(bars, probs_top3[::-1]):
210
  ax[1].text(prob + 1, bar.get_y() + bar.get_height()/2,
@@ -227,16 +131,4 @@ def classify_bytes(image_bytes: bytes):
227
  "top1": results[0],
228
  "top5": results,
229
  "plot": f"data:image/png;base64,{plot_b64}"
230
- }
231
-
232
-
233
- # ─────────────────────────────────────────────
234
- # LOCAL TEST
235
- # ─────────────────────────────────────────────
236
- if __name__ == "__main__":
237
- test_img = "sample3.jpg"
238
- with open(test_img, "rb") as f:
239
- img_bytes = f.read()
240
- out = classify_bytes(img_bytes)
241
- print(json.dumps(out["top5"], indent=2))
242
- print("\nPlot base64 length:", len(out["plot"]))
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  # app/inference.py
2
  import os
3
  import io
 
108
  edgecolor="none", height=0.6)
109
  ax[1].set_xlim(0, 100)
110
  ax[1].set_xlabel("Probability (%)", fontsize=11)
111
+ ax[1].set_title("Top-3 Predicted Species (DINOv2-G)", fontsize=12, weight="bold")
112
 
113
  for bar, prob in zip(bars, probs_top3[::-1]):
114
  ax[1].text(prob + 1, bar.get_y() + bar.get_height()/2,
 
131
  "top1": results[0],
132
  "top5": results,
133
  "plot": f"data:image/png;base64,{plot_b64}"
134
+ }
 
 
 
 
 
 
 
 
 
 
 
 
app/inference_yolo.py ADDED
@@ -0,0 +1,190 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import os
2
+ import io
3
+ import yaml
4
+ import base64
5
+ import torch
6
+ import numpy as np
7
+ import matplotlib.pyplot as plt
8
+ from PIL import Image
9
+ from ultralytics import YOLO
10
+ import matplotlib.patches as mpatches
11
+ # ─────────────────────────────────────────────
12
+ # CONFIG
13
+ # ─────────────────────────────────────────────
14
+ BASE_DIR = os.path.dirname(__file__)
15
+ MODEL_PATH = os.path.join(BASE_DIR, "best.pt") # Make sure to upload this!
16
+ DATA_YAML = os.path.join(BASE_DIR, "data.yaml") # Make sure to upload this!
17
+ IMG_SIZE = 512
18
+
19
+ _model = None
20
+ _class_names = None
21
+ _mapping = None
22
+
23
+ # ─────────────────────────────────────────────
24
+ # LOAD MODEL & LABELS
25
+ # ─────────────────────────────────────────────
26
+ def load_yolo_model():
27
+ global _model, _class_names, _mapping
28
+
29
+ if _model is not None:
30
+ return _model, _class_names, _mapping
31
+
32
+ print(f"🔁 Loading YOLOv11 model from: {MODEL_PATH}")
33
+
34
+ # 1. Load Model
35
+ _model = YOLO(MODEL_PATH)
36
+
37
+ # 2. Load Class Names from YAML
38
+ if os.path.exists(DATA_YAML):
39
+ with open(DATA_YAML, "r") as f:
40
+ data = yaml.safe_load(f)
41
+ _class_names = data["names"]
42
+ else:
43
+ print("⚠️ data.yaml not found! Using placeholder names.")
44
+ _class_names = {i: f"Species {i}" for i in range(1000)}
45
+
46
+ # 3. Build Mapping (YOLO index -> Species Index)
47
+ # Assumes YOLO classes are strings of IDs like "10", "45", etc.
48
+ yolo_names = _model.names
49
+ try:
50
+ _mapping = {yolo_idx: int(name_str) for yolo_idx, name_str in yolo_names.items()}
51
+ except ValueError:
52
+ print("⚠️ Warning: YOLO class names are not integers. Mapping might fail.")
53
+ _mapping = {k: k for k in yolo_names.keys()}
54
+
55
+ print("✅ YOLO Model loaded.")
56
+ return _model, _class_names, _mapping
57
+
58
+ # ─────────────────────────────────────────────
59
+ # CLASSIFY IMAGE BYTES (YOLO)
60
+ # ─────────────────────────────────────────────
61
+ def classify_yolo_bytes(image_bytes: bytes):
62
+ model, class_names, mapping = load_yolo_model()
63
+
64
+ # Load image
65
+ image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
66
+
67
+ # Inference
68
+ results = model.predict(source=image, imgsz=IMG_SIZE, verbose=False)
69
+ probs = results[0].probs.data.cpu().numpy()
70
+
71
+ # Get Top-5
72
+ top5_indices = np.argsort(probs)[-5:][::-1]
73
+ top5_probs = probs[top5_indices]
74
+
75
+ results_list = []
76
+ for rank, (idx, prob) in enumerate(zip(top5_indices, top5_probs), 1):
77
+ # Map internal YOLO index -> Folder Index -> Real Name
78
+ species_idx = mapping.get(idx, idx)
79
+ # Handle list vs dict for class_names
80
+ if isinstance(class_names, list):
81
+ label = class_names[species_idx] if species_idx < len(class_names) else f"Unknown ({species_idx})"
82
+ else:
83
+ label = class_names.get(species_idx, f"Unknown ({species_idx})")
84
+
85
+ results_list.append({
86
+ "rank": rank,
87
+ "id": int(species_idx),
88
+ "label": label,
89
+ "score": float(prob),
90
+ })
91
+ # ───────────────────────────────
92
+ # GENERATE PLOT (Subtle Rounded Bars)
93
+ # ───────────────────────────────
94
+ import matplotlib.patches as mpatches # <--- Make sure this is imported!
95
+
96
+ top3 = results_list[:3]
97
+ labels = [p["label"] for p in top3][::-1]
98
+ probs_top3 = [p["score"] * 100 for p in top3][::-1]
99
+
100
+ # 1. Setup Style and Figure
101
+ plt.style.use("seaborn-v0_8-white")
102
+ fig, ax = plt.subplots(1, 2, figsize=(10, 4.5), gridspec_kw={'width_ratios': [1, 2]})
103
+ fig.patch.set_facecolor('white')
104
+
105
+ # ───────────────────────────────
106
+ # LEFT PANEL: INPUT IMAGE
107
+ # ───────────────────────────────
108
+ ax[0].imshow(image)
109
+ ax[0].axis("off")
110
+ ax[0].set_title("Input Image", fontsize=12, fontweight='bold', color='#374151', pad=10)
111
+
112
+ # Subtle border for image
113
+ for spine in ax[0].spines.values():
114
+ spine.set_visible(True)
115
+ spine.set_color('#E5E7EB')
116
+ spine.set_linewidth(1)
117
+
118
+ # ───────────────────────────────
119
+ # RIGHT PANEL: SUBTLE ROUNDED BARS
120
+ # ───────────────────────────────
121
+ modern_colors = ['#4F46E5', '#10B981', '#F59E0B'][::-1]
122
+
123
+ bar_height = 0.6
124
+ y_positions = range(len(labels))
125
+
126
+ # Draw Subtle Rounded Rectangles
127
+ for i, (prob, color) in enumerate(zip(probs_top3, modern_colors)):
128
+ # We start x at -2.0 to "hide" the left rounded corners behind the axis
129
+ # width = prob + 2.0 to compensate for the negative start
130
+ # rounding_size=0.15 gives a subtle "app-like" corner
131
+ bbox = mpatches.FancyBboxPatch(
132
+ (-2.0, i - bar_height/2), # (x, y) bottom-left corner
133
+ prob + 2.0, # width
134
+ bar_height, # height
135
+ boxstyle="round,pad=0,rounding_size=0.15", # <--- CONTROL ROUNDNESS HERE
136
+ color=color,
137
+ ec="none",
138
+ mutation_aspect=0.3 # Flattens the curve slightly to look better on wide charts
139
+ )
140
+ ax[1].add_patch(bbox)
141
+
142
+ # Clean styling
143
+ ax[1].set_xlim(0, 100) # Clips the negative part, making left side flat
144
+ ax[1].set_ylim(-0.8, 2.8)
145
+
146
+ ax[1].spines['top'].set_visible(False)
147
+ ax[1].spines['right'].set_visible(False)
148
+ ax[1].spines['left'].set_visible(False)
149
+ ax[1].spines['bottom'].set_color('#D1D5DB')
150
+
151
+ # Subtle vertical grid
152
+ ax[1].grid(axis='x', linestyle=':', linewidth=0.8, color='#E5E7EB', alpha=0.8)
153
+ ax[1].set_axisbelow(True)
154
+
155
+ # Typography
156
+ ax[1].set_title(f"Top-3 Predictions (YOLOv11)", fontsize=13, fontweight='bold', color='#111827', pad=10)
157
+ ax[1].set_xlabel("Confidence Score (%)", fontsize=10, color='#6B7280', labelpad=8)
158
+
159
+ # Manually set Y-ticks to center of bars
160
+ ax[1].set_yticks(y_positions)
161
+ ax[1].set_yticklabels(labels, fontsize=11, color='#374151')
162
+ ax[1].tick_params(axis='y', length=0)
163
+ ax[1].tick_params(axis='x', colors='#6B7280', labelsize=9)
164
+
165
+ # Add Value Labels
166
+ for y, prob in zip(y_positions, probs_top3):
167
+ # Text inside bar if wide enough, outside if short
168
+ text_color = 'white' if prob > 20 else '#374151'
169
+ x_pos = prob - 3 if prob > 20 else prob + 1.5
170
+ ha_align = 'right' if prob > 20 else 'left'
171
+
172
+ ax[1].text(x_pos, y,
173
+ f"{prob:.1f}%",
174
+ va="center", ha=ha_align,
175
+ fontsize=10, fontweight='bold', color=text_color)
176
+
177
+ plt.tight_layout()
178
+ # Save to Base64
179
+ buf = io.BytesIO()
180
+ plt.savefig(buf, format="png", bbox_inches="tight")
181
+ plt.close(fig)
182
+ buf.seek(0)
183
+ plot_b64 = base64.b64encode(buf.read()).decode("utf-8")
184
+ buf.close()
185
+
186
+ return {
187
+ "top1": results_list[0],
188
+ "top5": results_list,
189
+ "plot": f"data:image/png;base64,{plot_b64}"
190
+ }
app/lionfish.jpg ADDED

Git LFS Details

  • SHA256: 9dcd8bda475d60905de82573e2c559dba10f0024250e32116110a9443c1c02e1
  • Pointer size: 130 Bytes
  • Size of remote file: 60.2 kB
app/main.py CHANGED
@@ -1,9 +1,12 @@
1
  # app/main.py
2
- from fastapi import FastAPI, UploadFile, File
3
  from fastapi.staticfiles import StaticFiles
4
  from fastapi.responses import HTMLResponse, JSONResponse
5
  from app.model import load_model, predict_from_bytes
6
  from app.inference import load_classification_model, classify_bytes
 
 
 
7
  import json, os
8
 
9
 
@@ -65,9 +68,18 @@ async def generate_attention(file: UploadFile = File(...)):
65
  return result
66
 
67
  @app.post("/classify")
68
- async def classify(file: UploadFile = File(...)):
 
 
 
69
  image_bytes = await file.read()
70
- return classify_bytes(image_bytes)
 
 
 
 
 
 
71
 
72
  @app.get("/api")
73
  def api_root():
 
1
  # app/main.py
2
+ from fastapi import FastAPI, UploadFile, File, Form
3
  from fastapi.staticfiles import StaticFiles
4
  from fastapi.responses import HTMLResponse, JSONResponse
5
  from app.model import load_model, predict_from_bytes
6
  from app.inference import load_classification_model, classify_bytes
7
+ from app.inference import load_classification_model, classify_bytes
8
+ from app.inference_yolo import classify_yolo_bytes, load_yolo_model
9
+ # from app.model import load_model, predict_pca_from_bytes
10
  import json, os
11
 
12
 
 
68
  return result
69
 
70
  @app.post("/classify")
71
+ async def classify(
72
+ file: UploadFile = File(...),
73
+ model: str = Form("dino") # <--- Read 'model' from FormData (default 'dino')
74
+ ):
75
  image_bytes = await file.read()
76
+
77
+ if model == "yolo":
78
+ print("🧠 Running YOLOv11 Inference...")
79
+ return classify_yolo_bytes(image_bytes)
80
+ else:
81
+ print("🦕 Running DINOv2 Inference...")
82
+ return classify_bytes(image_bytes)
83
 
84
  @app.get("/api")
85
  def api_root():
app/model.py CHANGED
@@ -9,32 +9,31 @@ import torchvision.transforms as pth_transforms
9
  from safetensors.torch import load_file
10
  import base64
11
  from io import BytesIO
12
- import matplotlib.cm as cm
 
 
 
13
 
14
  # --- Make sure Python can find dinov2 properly ---
15
- sys.path.append(os.path.dirname(__file__)) # adds 'app/' to sys.path
16
  from dinov2.models.vision_transformer import vit_large
17
- from huggingface_hub import hf_hub_download
18
-
19
 
20
  # --- Constants ---
21
- # CKPT_PATH = os.path.join(os.path.dirname(__file__), "ckpt_merged_large", "model.safetensors")
22
  CKPT_PATH = hf_hub_download(
23
- repo_id="Arew99/dinov2-costum", # <-- fixed spelling
24
  filename="model.safetensors"
25
  )
26
 
27
  PATCH_SIZE = 14
28
- IMAGE_SIZE = (800,800)
29
-
30
-
31
 
32
  # -------------------------------------------------------
33
  # Load model
34
  # -------------------------------------------------------
35
  def load_model():
36
  device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
37
- print(f"Loading model on device: {device}")
38
 
39
  model = vit_large(
40
  patch_size=PATCH_SIZE,
@@ -47,11 +46,8 @@ def load_model():
47
  # Load weights
48
  state_dict = load_file(CKPT_PATH)
49
  keys_list = list(state_dict.keys())
50
-
51
- # Handle "model." prefix if present
52
  if keys_list and "model." in keys_list[0]:
53
  state_dict = {k.replace("model.", ""): v for k, v in state_dict.items()}
54
- print("Removed 'model.' prefix from keys")
55
 
56
  model.load_state_dict(state_dict, strict=False)
57
  for p in model.parameters():
@@ -59,57 +55,95 @@ def load_model():
59
 
60
  model.eval()
61
  model.to(device)
62
- print("✅ Model loaded and ready.")
63
  return model, device
64
 
65
 
66
  # -------------------------------------------------------
67
- # Image preprocessing
68
  # -------------------------------------------------------
69
- def preprocess_image(image_bytes):
70
- """Load and transform image for DINOv2 model"""
71
- image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
72
-
 
 
 
 
 
 
 
 
 
73
  transform = pth_transforms.Compose([
74
- pth_transforms.Resize(IMAGE_SIZE),
75
  pth_transforms.ToTensor(),
76
- pth_transforms.Normalize((0.485, 0.456, 0.406),
77
- (0.229, 0.224, 0.225)),
78
  ])
79
- img = transform(image)
80
- w, h = img.shape[1] - img.shape[1] % PATCH_SIZE, img.shape[2] - img.shape[2] % PATCH_SIZE
81
- img = img[:, :w, :h].unsqueeze(0)
82
- return img, (w, h)
83
-
84
- def predict_from_bytes(model_device_tuple, image_bytes):
85
- model, device = model_device_tuple
86
- img, (w, h) = preprocess_image(image_bytes)
87
-
88
- w_featmap = img.shape[-2] // PATCH_SIZE
89
- h_featmap = img.shape[-1] // PATCH_SIZE
90
-
91
- x = img.to(device)
92
- attentions = model.get_last_self_attention(x)
93
- nh = attentions.shape[1] # number of heads
94
-
95
- # Reshape attention maps
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
96
  attentions = attentions[0, :, 0, 1:].reshape(nh, -1)
97
- attentions = attentions.reshape(nh, w_featmap, h_featmap)
 
 
98
  attentions = nn.functional.interpolate(
99
  attentions.unsqueeze(0),
100
  scale_factor=PATCH_SIZE,
101
  mode="nearest"
102
  )[0].cpu().numpy()
103
 
104
- # --- Normalize and visualize ---
105
  all_heads_base64 = []
106
- original_image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
107
 
108
  for i in range(nh):
109
  head_attn = attentions[i]
 
110
  head_norm = (head_attn - head_attn.min()) / (head_attn.max() - head_attn.min() + 1e-8)
 
 
111
  heatmap = (cm.viridis(head_norm)[:, :, :3] * 255).astype(np.uint8)
112
- heatmap_img = Image.fromarray(heatmap).resize(original_image.size, Image.BILINEAR)
113
 
114
  buf = BytesIO()
115
  heatmap_img.save(buf, format="PNG")
@@ -117,18 +151,10 @@ def predict_from_bytes(model_device_tuple, image_bytes):
117
  head_b64 = base64.b64encode(buf.getvalue()).decode("utf-8")
118
  all_heads_base64.append(head_b64)
119
 
120
- # --- Mean attention map ---
121
- mean_attention = np.mean(attentions, axis=0)
122
- mean_norm = (mean_attention - mean_attention.min()) / (mean_attention.max() - mean_attention.min() + 1e-8)
123
- heatmap = (cm.viridis(mean_norm)[:, :, :3] * 255).astype(np.uint8)
124
- mean_img = Image.fromarray(heatmap).resize(original_image.size, Image.BILINEAR)
125
-
126
- buf = BytesIO()
127
- mean_img.save(buf, format="PNG")
128
- buf.seek(0)
129
- mean_b64 = base64.b64encode(buf.getvalue()).decode("utf-8")
130
-
131
  return {
132
- "mean_attention_map": mean_b64,
133
- "head_attention_maps": all_heads_base64,
134
- }
 
9
  from safetensors.torch import load_file
10
  import base64
11
  from io import BytesIO
12
+ from sklearn.decomposition import PCA
13
+ from sklearn.preprocessing import StandardScaler
14
+ import matplotlib.cm as cm # Needed for attention head colormaps
15
+ from huggingface_hub import hf_hub_download
16
 
17
  # --- Make sure Python can find dinov2 properly ---
18
+ sys.path.append(os.path.dirname(__file__))
19
  from dinov2.models.vision_transformer import vit_large
 
 
20
 
21
  # --- Constants ---
 
22
  CKPT_PATH = hf_hub_download(
23
+ repo_id="Arew99/dinov2-costum",
24
  filename="model.safetensors"
25
  )
26
 
27
  PATCH_SIZE = 14
28
+ # We use a fixed height for consistency
29
+ INFERENCE_HEIGHT = 616
 
30
 
31
  # -------------------------------------------------------
32
  # Load model
33
  # -------------------------------------------------------
34
  def load_model():
35
  device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
36
+ print(f"Loading DINOv2 (PCA + Heads) on device: {device}")
37
 
38
  model = vit_large(
39
  patch_size=PATCH_SIZE,
 
46
  # Load weights
47
  state_dict = load_file(CKPT_PATH)
48
  keys_list = list(state_dict.keys())
 
 
49
  if keys_list and "model." in keys_list[0]:
50
  state_dict = {k.replace("model.", ""): v for k, v in state_dict.items()}
 
51
 
52
  model.load_state_dict(state_dict, strict=False)
53
  for p in model.parameters():
 
55
 
56
  model.eval()
57
  model.to(device)
 
58
  return model, device
59
 
60
 
61
  # -------------------------------------------------------
62
+ # HYBRID PREDICTION (PCA + HEADS)
63
  # -------------------------------------------------------
64
+ def predict_from_bytes(model_device_tuple, image_bytes):
65
+ model, device = model_device_tuple
66
+
67
+ # 1. Load & Preprocess
68
+ original_image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
69
+ orig_w, orig_h = original_image.size
70
+
71
+ # Resize keeping aspect ratio, but ensuring dimensions are divisible by PATCH_SIZE
72
+ scale = INFERENCE_HEIGHT / orig_h
73
+ new_w = int(orig_w * scale)
74
+ new_w = new_w - (new_w % PATCH_SIZE)
75
+ new_h = INFERENCE_HEIGHT - (INFERENCE_HEIGHT % PATCH_SIZE)
76
+
77
  transform = pth_transforms.Compose([
78
+ pth_transforms.Resize((new_h, new_w)),
79
  pth_transforms.ToTensor(),
80
+ pth_transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
 
81
  ])
82
+
83
+ img_tensor = transform(original_image).unsqueeze(0).to(device)
84
+
85
+ # Grid dimensions
86
+ w_featmap = new_w // PATCH_SIZE
87
+ h_featmap = new_h // PATCH_SIZE
88
+
89
+ # ─────────────────────────────────────────────────────────────
90
+ # PART A: PCA VISUALIZATION (Main Result)
91
+ # ─────────────────────────────────────────────────────────────
92
+ with torch.no_grad():
93
+ out = model.forward_features(img_tensor)
94
+ features = out["x_norm_patchtokens"][0].cpu().numpy()
95
+
96
+ # Apply PCA (3 Components -> RGB)
97
+ scaler = StandardScaler()
98
+ features_scaled = scaler.fit_transform(features)
99
+ pca = PCA(n_components=3)
100
+ pca_features = pca.fit_transform(features_scaled)
101
+
102
+ # Reshape and Normalize
103
+ pca_img = pca_features.reshape(h_featmap, w_featmap, 3)
104
+ pca_min, pca_max = pca_img.min(axis=(0, 1)), pca_img.max(axis=(0, 1))
105
+ pca_img = (pca_img - pca_min) / (pca_max - pca_min + 1e-8)
106
+ pca_uint8 = (pca_img * 255).astype(np.uint8)
107
+
108
+ # Resize to original
109
+ pca_pil = Image.fromarray(pca_uint8).resize((orig_w, orig_h), resample=Image.NEAREST)
110
+
111
+ buf = BytesIO()
112
+ pca_pil.save(buf, format="PNG")
113
+ buf.seek(0)
114
+ pca_b64 = base64.b64encode(buf.getvalue()).decode("utf-8")
115
+
116
+ # ─────────────────────────────────────────────────────────────
117
+ # PART B: ATTENTION HEADS (The 16 small plots)
118
+ # ─────────────────────────────────────────────────────────────
119
+ # Get raw attention weights from the last layer
120
+ with torch.no_grad():
121
+ attentions = model.get_last_self_attention(img_tensor)
122
+
123
+ nh = attentions.shape[1] # number of heads (usually 16 for ViT-Large)
124
+
125
+ # Process attention maps
126
+ # Shape: [1, heads, tokens, tokens] -> Extract CLS token attention [1, heads, 0, 1:]
127
  attentions = attentions[0, :, 0, 1:].reshape(nh, -1)
128
+ attentions = attentions.reshape(nh, h_featmap, w_featmap)
129
+
130
+ # Upsample to match patch size visually
131
  attentions = nn.functional.interpolate(
132
  attentions.unsqueeze(0),
133
  scale_factor=PATCH_SIZE,
134
  mode="nearest"
135
  )[0].cpu().numpy()
136
 
 
137
  all_heads_base64 = []
 
138
 
139
  for i in range(nh):
140
  head_attn = attentions[i]
141
+ # Normalize per head for better contrast
142
  head_norm = (head_attn - head_attn.min()) / (head_attn.max() - head_attn.min() + 1e-8)
143
+
144
+ # Apply colormap (Viridis)
145
  heatmap = (cm.viridis(head_norm)[:, :, :3] * 255).astype(np.uint8)
146
+ heatmap_img = Image.fromarray(heatmap).resize((orig_w, orig_h), Image.BILINEAR)
147
 
148
  buf = BytesIO()
149
  heatmap_img.save(buf, format="PNG")
 
151
  head_b64 = base64.b64encode(buf.getvalue()).decode("utf-8")
152
  all_heads_base64.append(head_b64)
153
 
154
+ # ─────────────────────────────────────────────────────────────
155
+ # RETURN BOTH
156
+ # ─────────────────────────────────────────────────────────────
 
 
 
 
 
 
 
 
157
  return {
158
+ "pca_image": pca_b64, # The Main "Rainbow" Image
159
+ "head_attention_maps": all_heads_base64 # The 16 Head Plots
160
+ }
app/sargus.jpeg ADDED

Git LFS Details

  • SHA256: d133999157ff8b60491a05c5e8d694e6b6ae7d1124605843955e5b504a1fdfa3
  • Pointer size: 129 Bytes
  • Size of remote file: 4.08 kB
app/static/.DS_Store CHANGED
Binary files a/app/static/.DS_Store and b/app/static/.DS_Store differ
 
app/static/Correctindex.html DELETED
@@ -1,345 +0,0 @@
1
- <!doctype html>
2
- <html lang="en">
3
- <head>
4
- <meta charset="utf-8" />
5
- <meta name="viewport" content="width=device-width, initial-scale=1" />
6
- <title>NEMO Tools</title>
7
-
8
- <!-- TailwindCSS CDN -->
9
- <script src="https://cdn.tailwindcss.com"></script>
10
- <style>
11
- body {
12
- background-image: url('static/background.jpg');
13
- background-size: cover;
14
- background-position: center;
15
- background-attachment: fixed;
16
- background-repeat: no-repeat;
17
- color: #f9fafb;
18
- }
19
- body::before {
20
- content: "";
21
- position: fixed;
22
- top: 0;
23
- left: 0;
24
- width: 100%;
25
- height: 100%;
26
- background: rgba(0, 10, 20, 0.3);
27
- z-index: -1;
28
- }
29
- .card {
30
- background: rgba(255, 255, 255, 0.12);
31
- backdrop-filter: blur(10px);
32
- border-radius: 20px;
33
- box-shadow: 0 4px 20px rgba(0, 0, 0, 0.2);
34
- color: #f1f1f1;
35
- }
36
- .nav-link {
37
- color: #e0e0e0;
38
- }
39
- .nav-link.active {
40
- color: #60a5fa;
41
- border-bottom: 2px solid #60a5fa;
42
- }
43
- </style>
44
- </head>
45
-
46
- <body class="bg-gray-100 min-h-screen">
47
- <header class="bg-white shadow-sm">
48
- <div class="max-w-6xl mx-auto px-4 py-4 flex items-center justify-between">
49
- <div class="flex items-center gap-3">
50
- <img src="/static/assets/logo.png" alt="NEMO logo" class="h-10 w-10 rounded-full shadow-sm" />
51
- <div>
52
- <h1 class="text-lg font-bold text-indigo-600">NEMO tools</h1>
53
- <p class="text-xs text-gray-400">DINOv2 visualisation sandbox</p>
54
- </div>
55
- </div>
56
-
57
- <nav class="flex gap-3">
58
- <button id="tab-research" class="tab-btn text-gray-500 hover:text-indigo-600 px-3 py-1 rounded-md text-sm font-medium" onclick="showTab('research')">
59
- Research
60
- </button>
61
- <button id="tab-people" class="tab-btn text-gray-500 hover:text-indigo-600 px-3 py-1 rounded-md text-sm font-medium" onclick="showTab('people')">
62
- People
63
- </button>
64
- <button id="tab-tools" class="tab-btn text-indigo-600 bg-indigo-50 px-3 py-1 rounded-md text-sm font-medium" onclick="showTab('tools')">
65
- Tools
66
- </button>
67
- </nav>
68
- </div>
69
- </header>
70
-
71
- <main class="max-w-6xl mx-auto px-4 py-8">
72
- <section id="page-research" class="hidden">
73
- <h2 class="text-2xl font-semibold text-gray-800 mb-4">Research</h2>
74
- <p class="text-gray-500 mb-4">We can list publications, datasets, and experiment notes here later.</p>
75
- <div class="bg-white rounded-xl shadow p-6 text-gray-400 text-sm">(placeholder) Add your papers, abstracts, or GitHub repos here.</div>
76
- </section>
77
-
78
- <section id="page-people" class="hidden">
79
- <h2 class="text-2xl font-semibold text-gray-800 mb-4">People</h2>
80
- <p class="text-gray-500 mb-4">We can add your name, collaborators, and links to profiles here.</p>
81
- <div class="grid gap-4 md:grid-cols-3">
82
- <div class="bg-white rounded-xl shadow p-5">
83
- <h3 class="font-semibold text-gray-700">You</h3>
84
- <p class="text-gray-400 text-sm">Project owner</p>
85
- </div>
86
- <div class="bg-white rounded-xl shadow p-5">
87
- <h3 class="font-semibold text-gray-700">To add</h3>
88
- <p class="text-gray-400 text-sm">Collaborators / advisors</p>
89
- </div>
90
- <div class="bg-white rounded-xl shadow p-5">
91
- <h3 class="font-semibold text-gray-700">Contact</h3>
92
- <p class="text-gray-400 text-sm">Add email / GitHub here</p>
93
- </div>
94
- </div>
95
- </section>
96
-
97
- <section id="page-tools">
98
- <div class="bg-white shadow-lg rounded-2xl p-8 w-full">
99
- <h2 class="text-2xl font-bold text-indigo-600 mb-6 flex items-center gap-2"><span>🧰 Tools</span></h2>
100
- <div class="flex gap-3 mb-6 border-b pb-2">
101
- <button id="sub-attention" class="subtab-btn text-indigo-600 font-medium border-b-2 border-indigo-600 pb-1" onclick="showSubTool('attention')">🧠 Attention Maps</button>
102
- <button id="sub-classification" class="subtab-btn text-gray-500 hover:text-indigo-600 pb-1" onclick="showSubTool('classification')">🔍 Run Classification</button>
103
- </div>
104
-
105
- <!-- Attention Tool -->
106
- <div id="tool-attention">
107
- <div class="flex flex-col items-center gap-4 mb-6 justify-center">
108
- <input id="file" type="file" accept="image/*"
109
- class="block w-full md:w-auto text-sm text-gray-600
110
- file:mr-4 file:py-2 file:px-4
111
- file:rounded-full file:border-0
112
- file:text-sm file:font-semibold
113
- file:bg-indigo-50 file:text-indigo-600
114
- hover:file:bg-indigo-100"
115
- onchange="onAttentionImageSelected()" />
116
- </div>
117
-
118
- <div id="attention-extra" class="hidden flex flex-col items-center gap-6">
119
- <button id="runButton" onclick="runActiveTool()" class="px-8 py-3 bg-indigo-600 text-white text-lg font-semibold rounded-full shadow-md hover:bg-indigo-700 transition">
120
- ▶️ Run Attention
121
- </button>
122
- <div class="flex flex-col md:flex-row justify-center items-start gap-6">
123
- <div class="flex flex-col items-center">
124
- <h4 class="text-gray-600 mb-2 font-medium">Original Image</h4>
125
- <img id="original" class="rounded-lg shadow-md max-w-xs hidden" />
126
- </div>
127
- <div class="flex flex-col items-center">
128
- <h4 class="text-gray-600 mb-2 font-medium">Mean Attention Map</h4>
129
- <img id="output" class="rounded-lg shadow-md max-w-xs hidden" />
130
- </div>
131
- </div>
132
- <div id="headsContainer" class="hidden mt-8">
133
- <h4 class="text-gray-600 mb-3 font-medium text-center">All Attention Heads</h4>
134
- <div id="headsGrid" class="flex flex-wrap justify-center gap-3"></div>
135
- </div>
136
- <p id="status" class="text-center text-gray-500 mt-2 text-sm"></p>
137
- </div>
138
- </div>
139
-
140
- <!-- Classification Tool -->
141
- <div id="tool-classification" class="hidden">
142
- <div class="flex flex-col items-center gap-4 mb-6 justify-center">
143
-
144
- <input id="cls-file" type="file" accept="image/*"
145
- class="block w-full md:w-auto text-sm text-gray-600
146
- file:mr-4 file:py-2 file:px-4
147
- file:rounded-full file:border-0
148
- file:text-sm file:font-semibold
149
- file:bg-indigo-50 file:text-indigo-600
150
- hover:file:bg-indigo-100" />
151
- <button id="cls-run" onclick="runClassification()" style="display:none;"
152
- class="px-8 py-3 bg-green-600 text-white text-lg font-semibold rounded-full shadow-md hover:bg-green-700 transition">
153
- ▶️ Run Classification
154
- </button>
155
- </div>
156
- <div id="cls-result" class="text-center text-gray-700 mt-4 text-lg font-medium"></div>
157
- </div>
158
- </div>
159
- </section>
160
- </main>
161
-
162
- <script>
163
- function showTab(name) {
164
- const tabs = ["research", "people", "tools"];
165
- tabs.forEach(t => {
166
- document.getElementById("page-" + t).classList.add("hidden");
167
- document.getElementById("tab-" + t).classList.remove("bg-indigo-50", "text-indigo-600");
168
- document.getElementById("tab-" + t).classList.add("text-gray-500");
169
- });
170
- document.getElementById("page-" + name).classList.remove("hidden");
171
- document.getElementById("tab-" + name).classList.add("bg-indigo-50", "text-indigo-600");
172
- document.getElementById("tab-" + name).classList.remove("text-gray-500");
173
- }
174
-
175
- let activeTool = "attention";
176
- const runButton = () => document.getElementById("runButton");
177
-
178
- function showSubTool(name) {
179
- const subs = ["attention", "classification"];
180
- subs.forEach(s => {
181
- document.getElementById("tool-" + s).classList.add("hidden");
182
- document.getElementById("sub-" + s).classList.remove("text-indigo-600", "font-medium", "border-b-2", "border-indigo-600");
183
- document.getElementById("sub-" + s).classList.add("text-gray-500");
184
- });
185
- document.getElementById("tool-" + name).classList.remove("hidden");
186
- document.getElementById("sub-" + name).classList.add("text-indigo-600", "font-medium", "border-b-2", "border-indigo-600");
187
- document.getElementById("sub-" + name).classList.remove("text-gray-500");
188
- activeTool = name;
189
- }
190
-
191
- async function runActiveTool() {
192
- const btn = runButton();
193
- btn.disabled = true;
194
- btn.textContent = "⌛️ Running...";
195
- btn.classList.add("opacity-70", "cursor-not-allowed");
196
- try {
197
- if (activeTool === "attention") await sendAttention();
198
- else await runClassification();
199
- } catch (err) {
200
- console.error(err);
201
- alert("❌ Error while running model: " + err.message);
202
- }
203
- btn.disabled = false;
204
- btn.classList.remove("opacity-70", "cursor-not-allowed");
205
- btn.textContent = activeTool === "attention" ? "▶️ Run Attention" : "▶️ Run Classification";
206
- }
207
-
208
- function onAttentionImageSelected() {
209
- const fileInput = document.getElementById("file");
210
- const extra = document.getElementById("attention-extra");
211
- const original = document.getElementById("original");
212
- if (fileInput.files.length > 0) {
213
- extra.classList.remove("hidden");
214
- const reader = new FileReader();
215
- reader.onload = e => {
216
- original.src = e.target.result;
217
- original.classList.remove("hidden");
218
- };
219
- reader.readAsDataURL(fileInput.files[0]);
220
- } else {
221
- extra.classList.add("hidden");
222
- original.classList.add("hidden");
223
- }
224
- }
225
-
226
- async function sendAttention() {
227
- const fileInput = document.getElementById("file");
228
- const file = fileInput.files[0];
229
- const output = document.getElementById("output");
230
- const status = document.getElementById("status");
231
- const headsContainer = document.getElementById("headsContainer");
232
- const headsGrid = document.getElementById("headsGrid");
233
- if (!file) {
234
- alert("Please choose an image first!");
235
- return;
236
- }
237
- output.classList.add("hidden");
238
- headsContainer.classList.add("hidden");
239
- headsGrid.innerHTML = "";
240
- status.textContent = "⌛️ Model is running...";
241
- const fd = new FormData();
242
- fd.append("file", file);
243
- try {
244
- const res = await fetch("/attention", { method: "POST", body: fd });
245
- if (!res.ok) throw new Error(`Server error: ${res.status}`);
246
- const json = await res.json();
247
- output.src = "data:image/png;base64," + json.mean_attention_map;
248
- output.classList.remove("hidden");
249
- if (json.head_attention_maps && json.head_attention_maps.length > 0) {
250
- json.head_attention_maps.forEach((headB64, i) => {
251
- const img = document.createElement("img");
252
- img.src = "data:image/png;base64," + headB64;
253
- img.alt = `Head ${i + 1}`;
254
- img.className = "rounded-md shadow-sm";
255
- img.style.width = "120px";
256
- img.style.transition = "transform 0.2s";
257
- img.onmouseenter = () => (img.style.transform = "scale(1.1)");
258
- img.onmouseleave = () => (img.style.transform = "scale(1)");
259
- headsGrid.appendChild(img);
260
- });
261
- headsContainer.classList.remove("hidden");
262
- }
263
- status.textContent = "✅ Done!";
264
- } catch (err) {
265
- console.error(err);
266
- status.textContent = "❌ Error: " + err.message;
267
- }
268
- }
269
-
270
- async function runClassification() {
271
- const fileInput = document.getElementById("cls-file");
272
- const file = fileInput.files[0];
273
- const result = document.getElementById("cls-result");
274
- const btn = document.getElementById("cls-run");
275
-
276
- if (!file) return alert("Please choose an image to classify!");
277
-
278
- // Disable button + show "Running..."
279
- btn.disabled = true;
280
- btn.textContent = "⏳ Running...";
281
- btn.classList.add("opacity-70", "cursor-not-allowed");
282
-
283
- result.textContent = "⌛️ Model is running...";
284
-
285
- const fd = new FormData();
286
- fd.append("file", file);
287
-
288
- try {
289
- const res = await fetch("/classify", { method: "POST", body: fd });
290
- if (!res.ok) throw new Error(`Server error: ${res.status}`);
291
- const json = await res.json();
292
-
293
- if (json.top5) {
294
- result.innerHTML = `
295
- <h3 class="font-semibold text-indigo-600 mb-2">Top-5 Predictions</h3>
296
- ${json.top5.map(p => `
297
- <div class="flex justify-between border-b py-1">
298
- <span>${p.rank}. ${p.label}</span>
299
- <span class="text-gray-500">${(p.score * 100).toFixed(2)}%</span>
300
- </div>
301
- `).join("")}
302
- `;
303
- } else {
304
- result.textContent = "No predictions returned.";
305
- }
306
-
307
- if (json.plot) {
308
- const plotImg = document.createElement("img");
309
- plotImg.src = json.plot;
310
- plotImg.alt = "Top-3 Predicted Species";
311
- plotImg.style.display = "block";
312
- plotImg.style.margin = "20px auto";
313
- plotImg.style.maxWidth = "800px";
314
- result.appendChild(plotImg);
315
- }
316
- } catch (err) {
317
- console.error(err);
318
- result.textContent = "❌ Error: " + err.message;
319
- } finally {
320
- // Restore button state
321
- btn.disabled = false;
322
- btn.classList.remove("opacity-70", "cursor-not-allowed");
323
- btn.textContent = "▶️ Run Classification";
324
- }
325
- }
326
-
327
-
328
- // Show classification button only after image is selected
329
- document.addEventListener("DOMContentLoaded", () => {
330
- const fileInput = document.getElementById("cls-file");
331
- const runBtn = document.getElementById("cls-run");
332
- fileInput.addEventListener("change", () => {
333
- if (fileInput.files && fileInput.files.length > 0) {
334
- runBtn.style.display = "inline-block";
335
- } else {
336
- runBtn.style.display = "none";
337
- }
338
- });
339
- });
340
-
341
- showTab("tools");
342
- showSubTool("attention");
343
- </script>
344
- </body>
345
- </html>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
app/static/background.jpg CHANGED

Git LFS Details

  • SHA256: 7ebfbdb3eef08b586d7ded5ed87a8e81f8a588f9bc3954c8e6b1df603956fe15
  • Pointer size: 130 Bytes
  • Size of remote file: 90.5 kB
app/static/index.html CHANGED
@@ -1,13 +1,26 @@
1
  <!doctype html>
2
  <html lang="en">
3
  <head>
 
 
 
 
 
 
 
 
4
  <meta charset="utf-8" />
5
  <meta name="viewport" content="width=device-width, initial-scale=1" />
6
  <title>NEMO Tools</title>
7
 
8
- <!-- TailwindCSS -->
9
  <script src="https://cdn.tailwindcss.com"></script>
10
  <style>
 
 
 
 
 
 
11
  body {
12
  background-image: url('/static/background.jpg');
13
  background-size: cover;
@@ -27,92 +40,490 @@
27
  </style>
28
  </head>
29
 
30
- <body class="bg-gray-100 min-h-screen">
31
 
32
- <!-- Header -->
33
- <header class="bg-white shadow-sm">
34
- <div class="max-w-6xl mx-auto px-4 py-4 flex items-center justify-between">
35
- <div class="flex items-center gap-3">
36
- <img src="/static/assets/logo.png" alt="NEMO logo" class="h-10 w-10 rounded-full shadow-sm" />
 
 
 
 
 
37
  <div>
38
- <h1 class="text-lg font-bold text-indigo-600">NEMO tools</h1>
39
- <p class="text-xs text-gray-400">DINOv2 visualisation sandbox</p>
40
  </div>
41
  </div>
42
- <nav class="flex gap-3">
43
- <button id="tab-research" class="tab-btn text-gray-500 hover:text-indigo-600 px-3 py-1 rounded-md text-sm font-medium" onclick="showTab('research')">Research</button>
44
- <button id="tab-people" class="tab-btn text-gray-500 hover:text-indigo-600 px-3 py-1 rounded-md text-sm font-medium" onclick="showTab('people')">People</button>
45
- <button id="tab-tools" class="tab-btn text-indigo-600 bg-indigo-50 px-3 py-1 rounded-md text-sm font-medium" onclick="showTab('tools')">Tools</button>
 
 
 
 
 
 
 
 
 
 
 
 
 
46
  </nav>
 
47
  </div>
48
  </header>
49
 
50
- <!-- Main Content -->
51
- <main class="max-w-6xl mx-auto px-4 py-8">
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
52
  <section id="page-tools">
53
  <div class="bg-white shadow-lg rounded-2xl p-8 w-full">
54
  <h2 class="text-2xl font-bold text-indigo-600 mb-6 flex items-center gap-2">🧰 Tools</h2>
55
 
56
- <!-- Subtabs -->
57
  <div class="flex gap-3 mb-6 border-b pb-2">
58
- <button id="sub-attention" class="subtab-btn text-indigo-600 font-medium border-b-2 border-indigo-600 pb-1" onclick="showSubTool('attention')">🧠 Mean Attention Map</button>
59
- <button id="sub-classification" class="subtab-btn text-gray-500 hover:text-indigo-600 pb-1" onclick="showSubTool('classification')">🔍 Run Classification</button>
60
  </div>
61
 
62
- <!-- 🧠 Attention Tool -->
63
- <div id="tool-attention">
64
- <div class="flex flex-col items-center gap-4 mb-6 justify-center">
65
- <div class="w-full max-w-md mx-auto text-center border-2 border-dashed border-gray-300 rounded-xl p-6 bg-gray-50 hover:bg-gray-100 transition">
66
- <p class="text-gray-700 font-semibold mb-2">Upload Image</p>
67
- <p class="text-gray-400 text-sm mb-4">Supported formats: JPG, PNG</p>
68
- <label for="file" class="inline-block bg-indigo-600 hover:bg-indigo-700 text-white font-semibold py-2 px-6 rounded-full cursor-pointer shadow-md transition">Choose File</label>
69
- <input id="file" type="file" accept="image/*" class="hidden" onchange="onAttentionFileSelected()" />
70
- <p id="attention-filename" class="text-gray-500 mt-3 text-sm"></p>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
71
  </div>
 
 
 
 
 
 
 
 
 
 
 
72
  </div>
73
 
74
- <div id="attention-extra" class="hidden flex flex-col items-center gap-6">
75
- <button id="runButton" onclick="runActiveTool()" class="px-8 py-3 bg-indigo-600 text-white text-lg font-semibold rounded-full shadow-md hover:bg-indigo-700 transition">▶️ Run Attention</button>
76
 
77
  <div class="flex flex-col md:flex-row justify-center items-start gap-6">
78
  <div class="flex flex-col items-center">
79
  <h4 class="text-gray-600 mb-2 font-medium">Original Image</h4>
80
- <img id="original" class="rounded-lg shadow-md max-w-xs hidden" />
81
  </div>
 
82
  <div class="flex flex-col items-center">
83
- <h4 class="text-gray-600 mb-2 font-medium">Mean Attention Map</h4>
84
- <img id="output" class="rounded-lg shadow-md max-w-xs hidden" />
85
  </div>
86
  </div>
87
 
88
- <div id="headsContainer" class="hidden mt-8">
89
  <h4 class="text-gray-600 mb-3 font-medium text-center">All Attention Heads</h4>
90
  <div id="headsGrid" class="flex flex-wrap justify-center gap-3"></div>
91
  </div>
 
92
  <p id="status" class="text-center text-gray-500 mt-2 text-sm"></p>
93
  </div>
 
94
  </div>
95
 
96
- <!-- 🔍 Classification Tool -->
97
- <div id="tool-classification" class="hidden flex flex-col items-center">
98
- <div class="w-full max-w-md mx-auto text-center border-2 border-dashed border-gray-300 rounded-xl p-6 bg-gray-50 hover:bg-gray-100 transition">
99
- <p class="text-gray-700 font-semibold mb-2">Upload Image</p>
100
- <p class="text-gray-400 text-sm mb-4">Supported formats: JPG, PNG</p>
101
- <label for="cls-file" class="inline-block bg-green-600 hover:bg-green-700 text-white font-semibold py-2 px-6 rounded-full cursor-pointer shadow-md transition">Choose File</label>
102
- <input id="cls-file" type="file" accept="image/*" class="hidden" onchange="onClsFileSelected()" />
103
- <p id="cls-filename" class="text-gray-500 mt-3 text-sm"></p>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
104
  </div>
105
 
106
- <button id="cls-run" onclick="runClassification()" style="display:none;" class="mt-5 px-8 py-3 bg-green-600 text-white text-lg font-semibold rounded-full shadow-md hover:bg-green-700 transition">▶️ Run Classification</button>
107
 
108
- <div id="cls-result" class="text-center text-gray-700 mt-6 text-lg font-medium"></div>
109
  </div>
 
110
  </div>
111
  </section>
112
  </main>
113
 
114
- <!-- Scripts -->
115
  <script>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
116
  let activeTool = "attention";
117
 
118
  function showSubTool(name) {
@@ -128,18 +539,25 @@
128
  activeTool = name;
129
  }
130
 
131
- // 🧠 Attention Tool
132
- function onAttentionFileSelected() {
133
- const fileInput = document.getElementById("file");
 
 
 
 
134
  const extra = document.getElementById("attention-extra");
135
  const original = document.getElementById("original");
136
- const nameEl = document.getElementById("attention-filename");
137
 
 
 
 
138
  if (fileInput.files.length > 0) {
139
- extra.classList.remove("hidden");
140
  const file = fileInput.files[0];
141
- nameEl.textContent = file.name;
 
142
 
 
143
  const reader = new FileReader();
144
  reader.onload = e => {
145
  original.src = e.target.result;
@@ -147,29 +565,45 @@
147
  };
148
  reader.readAsDataURL(file);
149
  } else {
150
- extra.classList.add("hidden");
151
- original.classList.add("hidden");
152
  nameEl.textContent = "";
 
153
  }
154
  }
155
 
156
- async function sendAttention() {
157
- const file = document.getElementById("file").files[0];
158
  const output = document.getElementById("output");
159
  const status = document.getElementById("status");
160
  const headsContainer = document.getElementById("headsContainer");
161
  const headsGrid = document.getElementById("headsGrid");
162
- const btn = document.getElementById("runButton");
 
163
 
164
  if (!file) return alert("Please choose an image first!");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
165
  output.classList.add("hidden");
166
- headsContainer.classList.add("hidden");
167
  headsGrid.innerHTML = "";
168
- // status.textContent = "⌛️ Model is running...";
169
 
170
- btn.disabled = true;
171
- btn.textContent = "⏳ Running...";
172
- btn.classList.add("opacity-70", "cursor-not-allowed");
 
173
 
174
  const fd = new FormData();
175
  fd.append("file", file);
@@ -177,32 +611,47 @@
177
  try {
178
  const res = await fetch("/attention", { method: "POST", body: fd });
179
  const json = await res.json();
180
- output.src = "data:image/png;base64," + json.mean_attention_map;
181
- output.classList.remove("hidden");
 
 
 
 
 
 
182
  if (json.head_attention_maps) {
183
  json.head_attention_maps.forEach((b64, i) => {
184
  const img = document.createElement("img");
185
  img.src = "data:image/png;base64," + b64;
186
- img.className = "rounded-md shadow-sm w-[120px]";
 
187
  headsGrid.appendChild(img);
188
  });
189
  headsContainer.classList.remove("hidden");
190
  }
191
- status.textContent = "✅ Done!";
 
 
192
  } catch (err) {
193
  status.textContent = "❌ Error: " + err.message;
194
  } finally {
 
195
  btn.disabled = false;
196
- btn.textContent = "▶️ Run Attention";
197
- btn.classList.remove("opacity-70", "cursor-not-allowed");
198
  }
199
  }
200
-
201
- // 🔍 Classification Tool
 
202
  function onClsFileSelected() {
203
  const fileInput = document.getElementById("cls-file");
204
  const fileName = document.getElementById("cls-filename");
205
  const runBtn = document.getElementById("cls-run");
 
 
 
 
206
  if (fileInput.files.length > 0) {
207
  fileName.textContent = fileInput.files[0].name;
208
  runBtn.style.display = "block";
@@ -219,48 +668,140 @@
219
 
220
  if (!file) return alert("Please choose an image to classify!");
221
 
 
222
  btn.disabled = true;
223
- btn.textContent = "⏳ Running...";
224
- btn.classList.add("opacity-70", "cursor-not-allowed");
225
- // result.textContent = "⌛️ Model is running...";
 
 
 
 
 
 
 
 
 
 
 
226
 
227
  const fd = new FormData();
228
  fd.append("file", file);
 
229
 
230
  try {
231
  const res = await fetch("/classify", { method: "POST", body: fd });
232
  const json = await res.json();
233
- if (json.top5) {
234
- result.innerHTML = `
235
- <h3 class="font-semibold text-indigo-600 mb-2">Top-5 Predictions</h3>
236
- ${json.top5.map(p => `
237
- <div class="flex justify-between border-b py-1">
238
- <span>${p.rank}. ${p.label}</span>
239
- <span class="text-gray-500">${(p.score * 100).toFixed(2)}%</span>
240
- </div>`).join("")}`;
241
- } else result.textContent = "No predictions returned.";
242
 
243
  if (json.plot) {
244
  const plotImg = document.createElement("img");
245
  plotImg.src = json.plot;
246
- plotImg.className = "block mx-auto mt-6 max-w-2xl";
247
  result.appendChild(plotImg);
 
 
248
  }
 
249
  } catch (err) {
250
  result.textContent = "❌ Error: " + err.message;
251
  } finally {
 
252
  btn.disabled = false;
253
- btn.textContent = "▶️ Run Classification";
254
- btn.classList.remove("opacity-70", "cursor-not-allowed");
255
  }
256
  }
257
 
 
258
  async function runActiveTool() {
259
- if (activeTool === "attention") await sendAttention();
260
  else await runClassification();
261
  }
262
 
263
- showSubTool("attention");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
264
  </script>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
265
  </body>
266
- </html>
 
1
  <!doctype html>
2
  <html lang="en">
3
  <head>
4
+ <script async src="https://www.googletagmanager.com/gtag/js?id=G-QFG67BD3FB"></script>
5
+ <script>
6
+ window.dataLayer = window.dataLayer || [];
7
+ function gtag(){dataLayer.push(arguments);}
8
+ gtag('js', new Date());
9
+ gtag('config', 'G-QFG67BD3FB');
10
+ </script>
11
+
12
  <meta charset="utf-8" />
13
  <meta name="viewport" content="width=device-width, initial-scale=1" />
14
  <title>NEMO Tools</title>
15
 
 
16
  <script src="https://cdn.tailwindcss.com"></script>
17
  <style>
18
+ .toggle-wrapper { display:flex; align-items:center; gap:10px; cursor:pointer; }
19
+ .toggle { width:44px; height:24px; background:#d1d5db; border-radius:9999px; position:relative; transition:0.2s; }
20
+ .toggle-circle { width:20px; height:20px; background:white; border-radius:50%; position:absolute; top:2px; left:2px; transition:0.2s; }
21
+ .toggle.active { background:#7c3aed; }
22
+ .toggle.active .toggle-circle { transform:translateX(20px); }
23
+
24
  body {
25
  background-image: url('/static/background.jpg');
26
  background-size: cover;
 
40
  </style>
41
  </head>
42
 
43
+ <body class="bg-gray-100 min-h-screen flex flex-col font-sans">
44
 
45
+ <header class="sticky top-0 z-50 backdrop-blur-md bg-white/70 border-b border-white/20 shadow-sm transition-all duration-300">
46
+ <div class="max-w-7xl mx-auto px-6 py-4 flex items-center justify-between">
47
+
48
+ <div class="flex items-center gap-4 group cursor-pointer">
49
+ <div class="relative">
50
+ <div class="absolute -inset-1 bg-gradient-to-r from-blue-400 to-indigo-500 rounded-full blur opacity-25 group-hover:opacity-50 transition duration-500"></div>
51
+ <img src="/static/assets/logo.png"
52
+ alt="NEMO logo"
53
+ class="relative h-16 w-16 rounded-full shadow-md border-2 border-white object-cover transform group-hover:scale-105 transition duration-300" />
54
+ </div>
55
  <div>
56
+ <h1 class="text-2xl font-extrabold text-indigo-900 tracking-tight leading-none group-hover:text-indigo-700 transition">NEMO tools</h1>
57
+ <p class="text-xs font-semibold text-indigo-500 uppercase tracking-wide mt-1">Mediterranean Fish Spenies Classification Tool</p>
58
  </div>
59
  </div>
60
+
61
+ <nav class="flex gap-2 md:gap-3 bg-white/50 p-1.5 rounded-full border border-white/40 shadow-inner">
62
+ <button id="tab-research"
63
+ class="tab-btn px-5 py-2 rounded-full text-sm font-semibold text-slate-600 hover:text-indigo-700 hover:bg-white/80 transition-all duration-200 focus:outline-none"
64
+ onclick="showTab('research')">
65
+ Research
66
+ </button>
67
+ <button id="tab-people"
68
+ class="tab-btn px-5 py-2 rounded-full text-sm font-semibold text-slate-600 hover:text-indigo-700 hover:bg-white/80 transition-all duration-200 focus:outline-none"
69
+ onclick="showTab('people')">
70
+ People
71
+ </button>
72
+ <button id="tab-tools"
73
+ class="tab-btn px-6 py-2 rounded-full text-sm font-bold text-white bg-indigo-600 shadow-md hover:bg-indigo-700 hover:shadow-lg transition-all duration-200 transform hover:-translate-y-0.5 focus:outline-none"
74
+ onclick="showTab('tools')">
75
+ Tools
76
+ </button>
77
  </nav>
78
+
79
  </div>
80
  </header>
81
 
82
+ <main class="max-w-6xl mx-auto px-4 py-8 w-full flex-grow">
83
+
84
+ <section id="page-people" class="hidden max-w-6xl mx-auto mb-12">
85
+ <div class="text-center mb-12">
86
+ <h2 class="text-3xl font-bold text-indigo-900">The Team</h2>
87
+ <p class="text-lg text-gray-800 mt-2">
88
+ Our interdisciplinary team combines expertise in ecology, marine biology, and artificial intelligence.
89
+ </p>
90
+ </div>
91
+ <div class="grid grid-cols-1 md:grid-cols-3 gap-8">
92
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
93
+ <img src="/static/team/kofidis.jpg" alt="Andreas Kofidis" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
94
+ <h3 class="text-lg font-bold text-gray-800">Andreas Kofidis</h3>
95
+ <span class="text-indigo-600 font-medium text-sm mb-3">Data Scientist, IACM-FORTH</span>
96
+ <p class="text-gray-600 text-sm leading-relaxed">
97
+ His research focuses on deep learning and computer vision, specifically in the development of robust algorithms for object detection and classification in challenging environments.
98
+ </p>
99
+ </div>
100
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
101
+ <img src="/static/team/dretaki.jpg" alt="Eleni Dretaki" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
102
+ <h3 class="text-lg font-bold text-gray-800">Eleni Dretaki</h3>
103
+ <span class="text-indigo-600 font-medium text-sm mb-3">Data Scientist, IACM-FORTH</span>
104
+ <p class="text-gray-600 text-sm leading-relaxed">
105
+ A data scientist specializing in applied mathematics and statistical analysis. Her work involves processing complex datasets and developing computational methodologies for scientific applications.
106
+ </p>
107
+ </div>
108
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
109
+ <img src="/static/team/zotou.jpg" alt="Maria Zotou" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
110
+ <h3 class="text-lg font-bold text-gray-800">Maria Zotou</h3>
111
+ <span class="text-indigo-600 font-medium text-sm mb-3">PhD candidate, UAegean</span>
112
+ <p class="text-gray-600 text-sm leading-relaxed">
113
+ Her research interests lie in marine biodiversity assessment and ecosystem monitoring. She specializes in the taxonomy of Mediterranean marine fauna and field survey methodologies.
114
+ </p>
115
+ </div>
116
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
117
+ <img src="/static/team/poursanidis.jpg" alt="Dimitris Poursanidis" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
118
+ <h3 class="text-lg font-bold text-gray-800">Dr. Dimitris Poursanidis</h3>
119
+ <span class="text-indigo-600 font-medium text-sm mb-3">Technical Researcher, IACM-FORTH</span>
120
+ <p class="text-gray-600 text-sm leading-relaxed">
121
+ An expert in remote sensing and coastal habitat mapping. His work utilizes satellite and drone imagery to study spatial patterns in marine environments and support conservation planning.
122
+ </p>
123
+ </div>
124
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
125
+ <img src="/static/team/doxa.jpg" alt="Aggeliki Doxa" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
126
+ <h3 class="text-lg font-bold text-gray-800">Dr. Aggeliki Doxa</h3>
127
+ <span class="text-indigo-600 font-medium text-sm mb-3">Associate Prof. of Biology, IACM-FORTH & UoC</span>
128
+ <p class="text-gray-600 text-sm leading-relaxed">
129
+ Her research focuses on ecological modeling, biodiversity dynamics, and the impact of environmental changes on species distributions and ecosystem services.
130
+ </p>
131
+ </div>
132
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
133
+ <img src="/static/team/mazaris.jpg" alt="Antonios Mazaris" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
134
+ <h3 class="text-lg font-bold text-gray-800">Prof. Antonios D. Mazaris</h3>
135
+ <span class="text-indigo-600 font-medium text-sm mb-3">Professor of Ecology, AUTH</span>
136
+ <p class="text-gray-600 text-sm leading-relaxed">
137
+ His work addresses large-scale conservation challenges, focusing on global biodiversity patterns, extinction risk assessments, and the effectiveness of protected area networks.
138
+ </p>
139
+ </div>
140
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
141
+ <img src="/static/team/katsanevakis.jpg" alt="Stelios Katsanevakis" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
142
+ <h3 class="text-lg font-bold text-gray-800">Prof. Stelios Katsanevakis</h3>
143
+ <span class="text-indigo-600 font-medium text-sm mb-3">Marine Ecologist, UAegean</span>
144
+ <p class="text-gray-600 text-sm leading-relaxed">
145
+ His research investigates marine ecosystem dynamics, biological invasions, and marine conservation planning, aiming to support ecosystem-based management strategies.
146
+ </p>
147
+ </div>
148
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
149
+ <img src="/static/team/kamarianakis.jpg" alt="Yiannis Kamarianakis" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
150
+ <h3 class="text-lg font-bold text-gray-800">Dr. Yiannis Kamarianakis</h3>
151
+ <span class="text-indigo-600 font-medium text-sm mb-3">Research Director, IACM-FORTH</span>
152
+ <p class="text-gray-600 text-sm leading-relaxed">
153
+ An expert in applied mathematics and statistics. He specializes in time-series analysis, spatial statistics, and the development of predictive models for complex systems.
154
+ </p>
155
+ </div>
156
+ <div class="bg-white rounded-2xl shadow-lg p-6 flex flex-col items-center text-center hover:shadow-xl transition">
157
+ <img src="/static/team/pantazis.jpg" alt="Yannis Pantazis" class="w-24 h-24 rounded-full mb-4 object-cover border-4 border-indigo-50 shadow-sm">
158
+ <h3 class="text-lg font-bold text-gray-800">Dr. Yannis Pantazis</h3>
159
+ <span class="text-indigo-600 font-medium text-sm mb-3">Principal Researcher, IACM-FORTH</span>
160
+ <p class="text-gray-600 text-sm leading-relaxed">
161
+ His research spans machine learning, signal processing, and data-driven discovery, focusing on developing novel AI architectures and mathematical frameworks for scientific applications.
162
+ </p>
163
+ </div>
164
+ </div>
165
+ </section>
166
+
167
+ <section id="page-research" class="hidden max-w-5xl mx-auto mb-12">
168
+ <div class="bg-white shadow-lg rounded-2xl p-8">
169
+
170
+ <div class="border-b pb-6 mb-6">
171
+ <h2 class="text-3xl font-bold text-indigo-900">NEMO-Tools Project</h2>
172
+ <p class="text-lg text-indigo-600 font-medium mt-2">
173
+ <span class="text-indigo-900 font-bold">NE</span>xt-generation
174
+ <span class="text-indigo-900 font-bold">MO</span>nitoring and mapping
175
+ <span class="text-indigo-900 font-bold">Tools</span> to assess marine ecosystems and biodiversity
176
+ </p>
177
+ </div>
178
+
179
+ <div class="prose max-w-none text-gray-700 leading-relaxed space-y-4">
180
+ <p>
181
+ <strong>NEMO-Tools</strong> is an ambitious research initiative designed to develop affordable, user-friendly, and smart tools for marine biodiversity monitoring. Its primary goal is to support the <strong>EU Biodiversity Strategy for 2030</strong> by providing high-precision data to measure ecosystem health and recovery.
182
+ </p>
183
+
184
+ <p>
185
+ While this web interface focuses on <strong>Artificial Intelligence (AI)</strong> for species identification, the full NEMO-Tools project integrates a hierarchical approach combining multiple cutting-edge technologies:
186
+ </p>
187
+
188
+ <div class="grid grid-cols-1 md:grid-cols-2 gap-6 my-8">
189
+ <div class="bg-indigo-50 p-5 rounded-xl border border-indigo-100">
190
+ <h3 class="font-bold text-indigo-800 mb-2 flex items-center gap-2">🤖 Artificial Intelligence</h3>
191
+ <p class="text-sm">
192
+ Utilizing Deep Learning (DINOv2, YOLOv11) to classify species from underwater imagery. This tool specifically utilizes the <strong>MEDFISH101</strong> dataset to identify 101 Mediterranean fish species.
193
+ </p>
194
+ </div>
195
+
196
+ <div class="bg-indigo-50 p-5 rounded-xl border border-indigo-100">
197
+ <h3 class="font-bold text-indigo-800 mb-2 flex items-center gap-2">🛰️ Remote Sensing</h3>
198
+ <p class="text-sm">
199
+ Employing satellite (Planet SuperDove) and drone (UAV) imagery to map macroalgae forests and detect marine megafauna in shallow coastal waters.
200
+ </p>
201
+ </div>
202
+
203
+ <div class="bg-indigo-50 p-5 rounded-xl border border-indigo-100">
204
+ <h3 class="font-bold text-indigo-800 mb-2 flex items-center gap-2">🧬 Environmental DNA (eDNA)</h3>
205
+ <p class="text-sm">
206
+ Developing cost-effective sampling protocols and low-cost 3D-printed probes to detect rare or elusive species through genetic material released in the water.
207
+ </p>
208
+ </div>
209
+
210
+ <div class="bg-indigo-50 p-5 rounded-xl border border-indigo-100">
211
+ <h3 class="font-bold text-indigo-800 mb-2 flex items-center gap-2">⚓ Robotics & Acoustics</h3>
212
+ <p class="text-sm">
213
+ Deploying upgraded ROVs for quantitative benthic assessments and Ground Unmanned Vehicles (GUVs) for sandy shore monitoring. Includes low-cost hydrophones for citizen-science acoustic monitoring.
214
+ </p>
215
+ </div>
216
+ </div>
217
+
218
+ <p>
219
+ The identification tool provided here addresses the challenge of monitoring visually similar species.
220
+ It leverages the <strong>MEDFISH101</strong> dataset, a rigorously curated collection of approximately
221
+ <strong>70,000 expert-validated images</strong>. This dataset represents 101 fish species
222
+ inhabiting the Hellenic and Mediterranean seas, spanning major families such as <em>Sparidae</em>,
223
+ <em>Serranidae</em>, and <em>Labridae</em>.
224
+ </p>
225
+ <p>
226
+ By encompassing commercial, protected, and invasive species—including the Indo-Pacific Lionfish (<em>Pterois miles</em>)—the tool facilitates near real-time ecological monitoring and early detection
227
+ of biological threats. Powered by state-of-the-art Vision Foundation Models achieving up to
228
+ <strong>94,12% accuracy</strong> using DINOv2-Giant, it empowers citizen scientists to contribute reliable data for
229
+ evidence-based marine management.
230
+ </p>
231
+ </div>
232
+
233
+ <div class="mt-10">
234
+ <h3 class="text-xl font-bold text-indigo-700 mb-4 flex items-center gap-2">
235
+ 🐟 MEDFISH101 Species List
236
+ </h3>
237
+ <p class="text-sm text-gray-500 mb-4">
238
+ The following 101 species are currently supported by the model classification engine:
239
+ </p>
240
+
241
+ <div class="bg-gray-50 rounded-xl p-6 border border-gray-200 h-[500px] overflow-y-auto">
242
+ <ul class="grid grid-cols-1 md:grid-cols-3 gap-y-2 gap-x-4 text-sm text-gray-700 font-medium">
243
+ <li><i>Aidablennius sphynx</i></li>
244
+ <li><i>Apogon imberbis</i></li>
245
+ <li><i>Arnoglossus laterna</i></li>
246
+ <li><i>Atherina boyeri</i></li>
247
+ <li><i>Atherina hepsetus</i></li>
248
+ <li><i>Boops boops</i></li>
249
+ <li><i>Bothus podas</i></li>
250
+ <li><i>Callionymus pusillus</i></li>
251
+ <li><i>Callionymus risso</i></li>
252
+ <li><i>Caranx crysos</i></li>
253
+ <li><i>Chelidonichthys lucerna</i></li>
254
+ <li><i>Chromis chromis</i></li>
255
+ <li><i>Coris julis</i></li>
256
+ <li><i>Dactylopterus volitans</i></li>
257
+ <li><i>Dasyatis pastinaca</i></li>
258
+ <li><i>Diplodus annularis</i></li>
259
+ <li><i>Diplodus puntazzo</i></li>
260
+ <li><i>Diplodus sargus</i></li>
261
+ <li><i>Diplodus vulgaris</i></li>
262
+ <li><i>Echiichthys vipera</i></li>
263
+ <li><i>Epinephelus costae</i></li>
264
+ <li><i>Epinephelus marginatus</i></li>
265
+ <li><i>Fistularia commersonii</i></li>
266
+ <li><i>Gobius auratus</i></li>
267
+ <li><i>Gobius bucchichi</i></li>
268
+ <li><i>Gobius cobitis</i></li>
269
+ <li><i>Gobius cruentatus</i></li>
270
+ <li><i>Gobius geniporus</i></li>
271
+ <li><i>Gobius incognitus</i></li>
272
+ <li><i>Gobius niger</i></li>
273
+ <li><i>Gobius paganellus</i></li>
274
+ <li><i>Gobius vittatus</i></li>
275
+ <li><i>Hippocampus guttulatus</i></li>
276
+ <li><i>Hippocampus hippocampus</i></li>
277
+ <li><i>Labrus bergylta</i></li>
278
+ <li><i>Labrus merula</i></li>
279
+ <li><i>Labrus mixtus</i></li>
280
+ <li><i>Labrus viridis</i></li>
281
+ <li><i>Lagocephalus sceleratus</i></li>
282
+ <li><i>Lithognathus mormyrus</i></li>
283
+ <li><i>Microlipophrys nigriceps</i></li>
284
+ <li><i>Mugil auratus</i></li>
285
+ <li><i>Mugil cephalus</i></li>
286
+ <li><i>Mullus barbatus</i></li>
287
+ <li><i>Mullus surmuletus</i></li>
288
+ <li><i>Muraena helena</i></li>
289
+ <li><i>Oblada melanura</i></li>
290
+ <li><i>Pagrus pagrus</i></li>
291
+ <li><i>Parablennius gattorugine</i></li>
292
+ <li><i>Parablennius incognitus</i></li>
293
+ <li><i>Parablennius pilicornis</i></li>
294
+ <li><i>Parablennius rouxi</i></li>
295
+ <li><i>Parablennius sanguinolentus</i></li>
296
+ <li><i>Parablennius tentacularis</i></li>
297
+ <li><i>Parablennius zvonimiri</i></li>
298
+ <li><i>Parupeneus forsskali</i></li>
299
+ <li><i>Phycis phycis</i></li>
300
+ <li><i>Plotosus lineatus</i></li>
301
+ <li><i>Pomatoschistus marmoratus</i></li>
302
+ <li><i>Pterois miles</i></li>
303
+ <li><i>Salaria pavo</i></li>
304
+ <li><i>Sargocentron rubrum</i></li>
305
+ <li><i>Sarpa salpa</i></li>
306
+ <li><i>Sciaena umbra</i></li>
307
+ <li><i>Scorpaena maderensis</i></li>
308
+ <li><i>Scorpaena notata</i></li>
309
+ <li><i>Scorpaena scrofa</i></li>
310
+ <li><i>Serranus cabrilla</i></li>
311
+ <li><i>Serranus hepatus</i></li>
312
+ <li><i>Serranus scriba</i></li>
313
+ <li><i>Siganus luridus</i></li>
314
+ <li><i>Siganus rivulatus</i></li>
315
+ <li><i>Solea solea</i></li>
316
+ <li><i>Sparisoma cretense</i></li>
317
+ <li><i>Sparus aurata</i></li>
318
+ <li><i>Spicara maena</i></li>
319
+ <li><i>Spicara smaris</i></li>
320
+ <li><i>Spondyliosoma cantharus</i></li>
321
+ <li><i>Stephanolepis diaspros</i></li>
322
+ <li><i>Symphodus cinereus</i></li>
323
+ <li><i>Symphodus mediterraneus</i></li>
324
+ <li><i>Symphodus melanocercus</i></li>
325
+ <li><i>Symphodus ocellatus</i></li>
326
+ <li><i>Symphodus roissali</i></li>
327
+ <li><i>Symphodus rostratus</i></li>
328
+ <li><i>Symphodus tinca</i></li>
329
+ <li><i>Syngnathus abaster</i></li>
330
+ <li><i>Syngnathus tenuirostris</i></li>
331
+ <li><i>Syngnathus typhle</i></li>
332
+ <li><i>Thalassoma pavo</i></li>
333
+ <li><i>Thorogobius ephippiatus</i></li>
334
+ <li><i>Torpedo marmorata</i></li>
335
+ <li><i>Torquigener flavimaculosus</i></li>
336
+ <li><i>Trachinus araneus</i></li>
337
+ <li><i>Trachinus draco</i></li>
338
+ <li><i>Trachinus radiatus</i></li>
339
+ <li><i>Tripterygion delaisi</i></li>
340
+ <li><i>Tripterygion melanurus</i></li>
341
+ <li><i>Tripterygion tripteronotus</i></li>
342
+ <li><i>Uranoscopus scaber</i></li>
343
+ <li><i>Xyrichtys novacula</i></li>
344
+ </ul>
345
+ </div>
346
+ </div>
347
+ </div>
348
+ </section>
349
+
350
  <section id="page-tools">
351
  <div class="bg-white shadow-lg rounded-2xl p-8 w-full">
352
  <h2 class="text-2xl font-bold text-indigo-600 mb-6 flex items-center gap-2">🧰 Tools</h2>
353
 
 
354
  <div class="flex gap-3 mb-6 border-b pb-2">
355
+ <button id="sub-classification" class="subtab-btn text-indigo-600 font-medium border-b-2 border-indigo-600 pb-1" onclick="showSubTool('classification')">🔍 Classification</button>
356
+ <button id="sub-attention" class="subtab-btn text-gray-500 hover:text-indigo-600 pb-1" onclick="showSubTool('attention')">🧠 Visualization</button>
357
  </div>
358
 
359
+ <div id="tool-attention" class="flex flex-col items-center w-full mt-6">
360
+
361
+ <div class="flex flex-col md:flex-row w-full items-stretch justify-center gap-8 px-4">
362
+
363
+ <div class="flex-1 flex flex-col justify-center items-start gap-4 min-w-[250px]">
364
+ <p class="text-gray-600 font-semibold mb-1 text-sm">Choose your model:</p>
365
+
366
+ <div class="toggle-wrapper group" onclick="setModel('dino')">
367
+ <div id="toggle-dino-vis" class="toggle active group-hover:ring-2 group-hover:ring-green-300">
368
+ <div class="toggle-circle shadow-sm"></div>
369
+ </div>
370
+ <div class="flex flex-col">
371
+ <span class="text-gray-800 font-bold text-sm">DINOv2-Giant</span>
372
+ <span class="text-gray-400 text-xs">High accuracy, slower</span>
373
+ </div>
374
+ </div>
375
+
376
+ <div class="toggle-wrapper group" onclick="setModel('yolo')">
377
+ <div id="toggle-yolo-vis" class="toggle group-hover:ring-2 group-hover:ring-green-300">
378
+ <div class="toggle-circle shadow-sm"></div>
379
+ </div>
380
+ <div class="flex flex-col">
381
+ <span class="text-gray-800 font-bold text-sm">YOLOv11-X</span>
382
+ <span class="text-gray-400 text-xs">Lower accuracy, faster</span>
383
+ </div>
384
+ </div>
385
+ </div>
386
+
387
+ <div class="flex-1 flex flex-col justify-center items-center text-center border-2 border-dashed border-gray-300 rounded-xl p-6 bg-gray-50 hover:bg-gray-100 transition min-w-[300px]">
388
+ <p class="text-gray-700 font-bold mb-2">Upload Image</p>
389
+ <p class="text-gray-400 text-xs mb-5">Supported: JPG, PNG</p>
390
+
391
+ <label for="vis-file"
392
+ class="inline-flex items-center justify-center gap-2 px-8 py-3 bg-green-600 hover:bg-green-700 text-white font-bold rounded-full cursor-pointer shadow-md transition-all transform hover:scale-105">
393
+ <svg xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" class="w-5 h-5">
394
+ <path stroke-linecap="round" stroke-linejoin="round" d="M3 16.5v2.25A2.25 2.25 0 005.25 21h13.5A2.25 2.25 0 0021 18.75V16.5m-13.5-9L12 3m0 0l4.5 4.5M12 3v13.5" />
395
+ </svg>
396
+ Choose File
397
+ </label>
398
+
399
+ <input id="vis-file" type="file" accept="image/*" class="hidden" onchange="onVisFileSelected()" />
400
+ <p id="vis-filename" class="text-gray-500 mt-3 text-xs italic font-medium"></p>
401
  </div>
402
+
403
+ <div class="flex-1 min-w-[250px] hidden md:block"></div>
404
+
405
+ </div>
406
+
407
+ <div class="w-full flex justify-center mt-6">
408
+ <button id="vis-run" onclick="runVisualization()"
409
+ style="display:none;"
410
+ class="px-8 py-3 bg-indigo-600 text-white text-lg font-semibold rounded-full shadow-md hover:bg-indigo-700 transition transform hover:scale-105">
411
+ ▶️ Run Visualization
412
+ </button>
413
  </div>
414
 
415
+ <div id="attention-extra" class="hidden flex flex-col items-center gap-6 w-full mt-8">
 
416
 
417
  <div class="flex flex-col md:flex-row justify-center items-start gap-6">
418
  <div class="flex flex-col items-center">
419
  <h4 class="text-gray-600 mb-2 font-medium">Original Image</h4>
420
+ <img id="original" class="rounded-lg shadow-md max-w-[200px] hidden" />
421
  </div>
422
+
423
  <div class="flex flex-col items-center">
424
+ <h4 class="text-gray-600 mb-2 font-medium">PCA Attention</h4>
425
+ <img id="output" class="rounded-lg shadow-md max-w-[200px] hidden" />
426
  </div>
427
  </div>
428
 
429
+ <div id="headsContainer" class="hidden mt-8 w-full">
430
  <h4 class="text-gray-600 mb-3 font-medium text-center">All Attention Heads</h4>
431
  <div id="headsGrid" class="flex flex-wrap justify-center gap-3"></div>
432
  </div>
433
+
434
  <p id="status" class="text-center text-gray-500 mt-2 text-sm"></p>
435
  </div>
436
+
437
  </div>
438
 
439
+ <div id="tool-classification" class="hidden flex flex-col items-start mt-6 w-full">
440
+
441
+ <div class="flex flex-col md:flex-row w-full items-stretch justify-center gap-8 px-4">
442
+
443
+ <div class="flex-1 flex flex-col justify-center items-start gap-4 min-w-[250px]">
444
+ <p class="text-gray-600 font-semibold mb-1 text-sm">Choose your model:</p>
445
+
446
+ <div class="toggle-wrapper group" onclick="setModel('dino')">
447
+ <div id="toggle-dino-cls" class="toggle active group-hover:ring-2 group-hover:ring-green-300">
448
+ <div class="toggle-circle shadow-sm"></div>
449
+ </div>
450
+ <div class="flex flex-col">
451
+ <span class="text-gray-800 font-bold text-sm">DINOv2-Giant</span>
452
+ <span class="text-gray-400 text-xs">High accuracy, slower</span>
453
+ </div>
454
+ </div>
455
+
456
+ <div class="toggle-wrapper group" onclick="setModel('yolo')">
457
+ <div id="toggle-yolo-cls" class="toggle group-hover:ring-2 group-hover:ring-green-300">
458
+ <div class="toggle-circle shadow-sm"></div>
459
+ </div>
460
+ <div class="flex flex-col">
461
+ <span class="text-gray-800 font-bold text-sm">YOLOv11-X</span>
462
+ <span class="text-gray-400 text-xs">Lower accuracy, faster</span>
463
+ </div>
464
+ </div>
465
+ </div>
466
+
467
+ <div class="flex-1 flex flex-col justify-center items-center text-center border-2 border-dashed border-gray-300 rounded-xl p-6 bg-gray-50 hover:bg-gray-100 transition min-w-[300px]">
468
+ <p class="text-gray-700 font-bold mb-2">Upload Image</p>
469
+ <p class="text-gray-400 text-xs mb-5">Supported: JPG, PNG</p>
470
+
471
+ <label for="cls-file"
472
+ class="inline-flex items-center justify-center gap-2 px-8 py-3 bg-green-600 hover:bg-green-700 text-white font-bold rounded-full cursor-pointer shadow-md transition-all transform hover:scale-105">
473
+ <svg xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="2" stroke="currentColor" class="w-5 h-5">
474
+ <path stroke-linecap="round" stroke-linejoin="round" d="M3 16.5v2.25A2.25 2.25 0 005.25 21h13.5A2.25 2.25 0 0021 18.75V16.5m-13.5-9L12 3m0 0l4.5 4.5M12 3v13.5" />
475
+ </svg>
476
+ Choose File
477
+ </label>
478
+
479
+ <input id="cls-file" type="file" accept="image/*" class="hidden" onchange="onClsFileSelected()" />
480
+ <p id="cls-filename" class="text-gray-500 mt-3 text-xs italic"></p>
481
+ </div>
482
+
483
+ <div class="flex-1 flex flex-col justify-center gap-4 min-w-[250px]">
484
+ <div>
485
+ <label class="block text-gray-700 font-bold mb-1 text-sm">Location of observation:</label>
486
+ <input id="obs-location" type="text" class="w-full border border-gray-300 rounded-lg px-3 py-2 text-sm text-gray-900 focus:ring-2 focus:ring-green-500 focus:border-green-500 transition-all focus:outline-none shadow-sm" placeholder="e.g. Aegean Sea" />
487
+ </div>
488
+ <div>
489
+ <label class="block text-gray-700 font-bold mb-1 text-sm">Potential species name:</label>
490
+ <input id="obs-species" type="text" class="w-full border border-gray-300 rounded-lg px-3 py-2 text-sm text-gray-900 focus:ring-2 focus:ring-green-500 focus:border-green-500 transition-all focus:outline-none shadow-sm" placeholder="e.g. Sparus aurata" />
491
+ </div>
492
+ </div>
493
+
494
+ </div>
495
+
496
+ <div class="w-full flex justify-center mt-6">
497
+ <button id="cls-run" onclick="runClassification()" style="display:none;" class="px-8 py-3 bg-green-600 text-white text-lg font-semibold rounded-full shadow-md hover:bg-green-700 transition">
498
+ ▶️ Run Classification
499
+ </button>
500
  </div>
501
 
502
+ <div id="cls-result" class="text-center text-gray-700 mt-6 text-lg font-medium w-full"></div>
503
 
 
504
  </div>
505
+
506
  </div>
507
  </section>
508
  </main>
509
 
 
510
  <script>
511
+ let selectedModel = "dino"; // default model
512
+
513
+ function setModel(name) {
514
+ selectedModel = name;
515
+
516
+ // Update Visualization Toggles
517
+ document.getElementById("toggle-dino-vis")?.classList.toggle("active", name === "dino");
518
+ document.getElementById("toggle-yolo-vis")?.classList.toggle("active", name === "yolo");
519
+
520
+ // Update Classification Toggles
521
+ document.getElementById("toggle-dino-cls")?.classList.toggle("active", name === "dino");
522
+ document.getElementById("toggle-yolo-cls")?.classList.toggle("active", name === "yolo");
523
+
524
+ console.log("MODEL SELECTED:", selectedModel);
525
+ }
526
+
527
  let activeTool = "attention";
528
 
529
  function showSubTool(name) {
 
539
  activeTool = name;
540
  }
541
 
542
+ // ──────────────────────────────────────────────
543
+ // 🧠 VISUALIZATION LOGIC
544
+ // ──────────────────────────────────────────────
545
+ function onVisFileSelected() {
546
+ const fileInput = document.getElementById("vis-file");
547
+ const nameEl = document.getElementById("vis-filename");
548
+ const runBtn = document.getElementById("vis-run");
549
  const extra = document.getElementById("attention-extra");
550
  const original = document.getElementById("original");
 
551
 
552
+ // Reset previous results
553
+ extra.classList.add("hidden");
554
+
555
  if (fileInput.files.length > 0) {
 
556
  const file = fileInput.files[0];
557
+ nameEl.textContent = file.name; // Show filename
558
+ runBtn.style.display = "block"; // Show Run button
559
 
560
+ // Pre-load original image for display later
561
  const reader = new FileReader();
562
  reader.onload = e => {
563
  original.src = e.target.result;
 
565
  };
566
  reader.readAsDataURL(file);
567
  } else {
 
 
568
  nameEl.textContent = "";
569
+ runBtn.style.display = "none";
570
  }
571
  }
572
 
573
+ async function runVisualization() {
574
+ const file = document.getElementById("vis-file").files[0];
575
  const output = document.getElementById("output");
576
  const status = document.getElementById("status");
577
  const headsContainer = document.getElementById("headsContainer");
578
  const headsGrid = document.getElementById("headsGrid");
579
+ const btn = document.getElementById("vis-run");
580
+ const extra = document.getElementById("attention-extra");
581
 
582
  if (!file) return alert("Please choose an image first!");
583
+
584
+ // 1. Change Button State to Loading
585
+ btn.disabled = true;
586
+ btn.classList.add("opacity-80", "cursor-not-allowed");
587
+ // Replace text with spinner
588
+ btn.innerHTML = `
589
+ <svg class="animate-spin -ml-1 mr-3 h-5 w-5 text-white inline-block" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24">
590
+ <circle class="opacity-25" cx="12" cy="12" r="10" stroke="currentColor" stroke-width="4"></circle>
591
+ <path class="opacity-75" fill="currentColor" d="M4 12a8 8 0 018-8V0C5.373 0 0 5.373 0 12h4zm2 5.291A7.962 7.962 0 014 12H0c0 3.042 1.135 5.824 3 7.938l3-2.647z"></path>
592
+ </svg>
593
+ Processing...
594
+ `;
595
+
596
+ // Setup UI for results
597
+ extra.classList.remove("hidden");
598
  output.classList.add("hidden");
599
+ headsContainer.classList.add("hidden");
600
  headsGrid.innerHTML = "";
601
+ status.textContent = "Running model inference...";
602
 
603
+ // Google Analytics
604
+ if (typeof gtag === 'function') {
605
+ gtag('event', 'run_visualization', { 'event_category': 'Model', 'event_label': 'DINOv2_Hybrid' });
606
+ }
607
 
608
  const fd = new FormData();
609
  fd.append("file", file);
 
611
  try {
612
  const res = await fetch("/attention", { method: "POST", body: fd });
613
  const json = await res.json();
614
+
615
+ // 1. Show PCA Image
616
+ if (json.pca_image) {
617
+ output.src = "data:image/png;base64," + json.pca_image;
618
+ output.classList.remove("hidden");
619
+ }
620
+
621
+ // 2. Show 16 Attention Heads
622
  if (json.head_attention_maps) {
623
  json.head_attention_maps.forEach((b64, i) => {
624
  const img = document.createElement("img");
625
  img.src = "data:image/png;base64," + b64;
626
+ img.className = "rounded-md shadow-sm w-[120px] transition hover:scale-110";
627
+ img.title = `Head ${i+1}`;
628
  headsGrid.appendChild(img);
629
  });
630
  headsContainer.classList.remove("hidden");
631
  }
632
+
633
+ status.textContent = "✅ Visualization Complete!";
634
+
635
  } catch (err) {
636
  status.textContent = "❌ Error: " + err.message;
637
  } finally {
638
+ // 2. Reset Button State
639
  btn.disabled = false;
640
+ btn.innerHTML = "▶️ Run Visualization";
641
+ btn.classList.remove("opacity-80", "cursor-not-allowed");
642
  }
643
  }
644
+ // ──────────────────────────────────────────────
645
+ // 🔍 CLASSIFICATION LOGIC
646
+ // ──────────────────────────────────────────────
647
  function onClsFileSelected() {
648
  const fileInput = document.getElementById("cls-file");
649
  const fileName = document.getElementById("cls-filename");
650
  const runBtn = document.getElementById("cls-run");
651
+
652
+ // Clear previous results
653
+ document.getElementById("cls-result").innerHTML = "";
654
+
655
  if (fileInput.files.length > 0) {
656
  fileName.textContent = fileInput.files[0].name;
657
  runBtn.style.display = "block";
 
668
 
669
  if (!file) return alert("Please choose an image to classify!");
670
 
671
+ // 1. Change Button State to Loading
672
  btn.disabled = true;
673
+ btn.classList.add("opacity-80", "cursor-not-allowed");
674
+ // Replace text with spinner
675
+ btn.innerHTML = `
676
+ <svg class="animate-spin -ml-1 mr-3 h-5 w-5 text-white inline-block" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24">
677
+ <circle class="opacity-25" cx="12" cy="12" r="10" stroke="currentColor" stroke-width="4"></circle>
678
+ <path class="opacity-75" fill="currentColor" d="M4 12a8 8 0 018-8V0C5.373 0 0 5.373 0 12h4zm2 5.291A7.962 7.962 0 014 12H0c0 3.042 1.135 5.824 3 7.938l3-2.647z"></path>
679
+ </svg>
680
+ Processing...
681
+ `;
682
+
683
+ // Analytics
684
+ if (typeof gtag === 'function') {
685
+ gtag('event', 'run_classification', { 'event_category': 'Model', 'event_label': selectedModel });
686
+ }
687
 
688
  const fd = new FormData();
689
  fd.append("file", file);
690
+ fd.append("model", selectedModel);
691
 
692
  try {
693
  const res = await fetch("/classify", { method: "POST", body: fd });
694
  const json = await res.json();
695
+
696
+ if (json.top5) result.innerHTML = ""; // Clear previous results
 
 
 
 
 
 
 
697
 
698
  if (json.plot) {
699
  const plotImg = document.createElement("img");
700
  plotImg.src = json.plot;
701
+ plotImg.className = "block mx-auto mt-6 max-w-4xl rounded-lg shadow-md";
702
  result.appendChild(plotImg);
703
+ } else {
704
+ result.textContent = "No predictions returned.";
705
  }
706
+
707
  } catch (err) {
708
  result.textContent = "❌ Error: " + err.message;
709
  } finally {
710
+ // 2. Reset Button State
711
  btn.disabled = false;
712
+ btn.innerHTML = "▶️ Run Classification";
713
+ btn.classList.remove("opacity-80", "cursor-not-allowed");
714
  }
715
  }
716
 
717
+ // Helper for "Run Active Tool" if needed
718
  async function runActiveTool() {
719
+ if (activeTool === "attention") await runVisualization();
720
  else await runClassification();
721
  }
722
 
723
+ // TAB SWITCHING LOGIC
724
+ // TAB SWITCHING LOGIC (Fixed)
725
+ function showTab(tabName) {
726
+ // 1. Hide all pages
727
+ ['tools', 'research', 'people'].forEach(page => {
728
+ const el = document.getElementById('page-' + page);
729
+ if(el) el.classList.add('hidden');
730
+ });
731
+
732
+ // 2. Show selected page
733
+ const selected = document.getElementById('page-' + tabName);
734
+ if(selected) selected.classList.remove('hidden');
735
+
736
+ // 3. Reset All Buttons to "Inactive" State
737
+ const allButtons = document.querySelectorAll('.tab-btn');
738
+ allButtons.forEach(btn => {
739
+ // Remove the "Active" styles (Purple/Indigo background & White text)
740
+ btn.classList.remove('bg-indigo-600', 'text-white', 'shadow-md');
741
+
742
+ // Add the "Inactive" styles (Transparent background & Gray text)
743
+ btn.classList.add('text-slate-600', 'hover:bg-white/80', 'hover:text-indigo-700');
744
+ });
745
+
746
+ // 4. Set Active Button Style
747
+ const activeBtn = document.getElementById('tab-' + tabName);
748
+ if(activeBtn) {
749
+ // Remove "Inactive" styles
750
+ activeBtn.classList.remove('text-slate-600', 'hover:bg-white/80', 'hover:text-indigo-700');
751
+
752
+ // Add "Active" styles (Purple/Indigo background & White text)
753
+ activeBtn.classList.add('bg-indigo-600', 'text-white', 'shadow-md');
754
+ }
755
+ }
756
+
757
+ // Initialize
758
+ showSubTool("classification");
759
  </script>
760
+
761
+ <footer class="bg-slate-900 border-t border-slate-800 mt-20 pb-12 text-white">
762
+ <div class="max-w-7xl mx-auto px-6 pt-16">
763
+
764
+ <div class="text-center mb-10">
765
+ <h3 class="text-sm font-bold tracking-[0.2em] text-slate-400 uppercase">
766
+ Supported By
767
+ </h3>
768
+ <div class="w-16 h-1 bg-indigo-500 mx-auto mt-4 rounded-full shadow-lg shadow-indigo-500/50"></div>
769
+ </div>
770
+
771
+ <div class="bg-white rounded-3xl p-10 shadow-2xl mx-auto max-w-6xl">
772
+ <div class="flex flex-wrap justify-center items-center gap-x-16 gap-y-12">
773
+
774
+ <a href="https://www.iacm.forth.gr/" target="_blank" rel="noopener noreferrer">
775
+ <img src="/static/logos/forth.png" alt="IACM-FORTH" class="h-24 md:h-32 w-auto object-contain transition-transform duration-300 hover:scale-110 cursor-pointer">
776
+ </a>
777
+
778
+ <a href="https://www.aegean.gr/" target="_blank" rel="noopener noreferrer">
779
+ <img src="/static/logos/aegean.png" alt="University of the Aegean" class="h-24 md:h-32 w-auto object-contain transition-transform duration-300 hover:scale-110 cursor-pointer">
780
+ </a>
781
+
782
+ <a href="https://www.elidek.gr/en/homepage/" target="_blank" rel="noopener noreferrer">
783
+ <img src="/static/logos/hfri.png" alt="HFRI" class="h-24 md:h-32 w-auto object-contain transition-transform duration-300 hover:scale-110 cursor-pointer">
784
+ </a>
785
+
786
+ <a href="https://www.uoc.gr/en/home/" target="_blank" rel="noopener noreferrer">
787
+ <img src="/static/logos/uoc.png" alt="University of Crete" class="h-24 md:h-32 w-auto object-contain transition-transform duration-300 hover:scale-110 cursor-pointer">
788
+ </a>
789
+
790
+ <a href="https://www.auth.gr/en/" target="_blank" rel="noopener noreferrer">
791
+ <img src="/static/logos/auth.png" alt="Aristotle University of Thessaloniki" class="h-12 md:h-14 w-auto object-contain transition-transform duration-300 hover:scale-110 cursor-pointer">
792
+ </a>
793
+
794
+ </div>
795
+ </div>
796
+
797
+ <div class="mt-16 text-center border-t border-slate-800 pt-8">
798
+ <p class="text-slate-500 text-sm">
799
+ © 2025 NEMO-Tools. Mediterranean Fish Species Classification. All rights reserved.
800
+ </p>
801
+ </div>
802
+
803
+ </div>
804
+ </footer>
805
+
806
  </body>
807
+ </html>
app/static/logos/aegean.png ADDED

Git LFS Details

  • SHA256: 6e4dbc822b12d9df38654309b7652b4e4077630c24b32988fe7b0d9823b9fa64
  • Pointer size: 130 Bytes
  • Size of remote file: 10 kB
app/static/logos/auth.png ADDED

Git LFS Details

  • SHA256: eeb7576f6bffdc4f4b2148968626a97dcc4dd42302739125c2c23256a4d87294
  • Pointer size: 130 Bytes
  • Size of remote file: 52.8 kB
app/static/logos/forth.png ADDED

Git LFS Details

  • SHA256: 8fa00fb8fb57f58cdc3ae82fd38d3299825387edc657e0467022cc3e6d2849ff
  • Pointer size: 131 Bytes
  • Size of remote file: 180 kB
app/static/logos/hfri.png ADDED

Git LFS Details

  • SHA256: 28cca9dfe52067ab0b399591301b815e8ae62a3cdafaded30a5d6eabba9acf75
  • Pointer size: 130 Bytes
  • Size of remote file: 27.7 kB
app/static/logos/uoc.png ADDED

Git LFS Details

  • SHA256: 2d29f9539c403b23c1bf3efe383c5353ceb13c65afe79ab6b239700a1b20212d
  • Pointer size: 131 Bytes
  • Size of remote file: 493 kB
app/static/team/.DS_Store ADDED
Binary file (6.15 kB). View file
 
app/static/team/doxa.jpg ADDED

Git LFS Details

  • SHA256: 5aa05b1ac788d3f2b942fd65f211452f3ffdbd65030116f4a777045464bb73ba
  • Pointer size: 129 Bytes
  • Size of remote file: 7.92 kB
app/static/team/dretaki.jpg ADDED

Git LFS Details

  • SHA256: f7a1fe04a3e3f2156804924c168b8ffa988c4745f6a9f3d70a95812f2971b704
  • Pointer size: 130 Bytes
  • Size of remote file: 44.8 kB
app/static/team/kamarianakis.jpg ADDED

Git LFS Details

  • SHA256: 323d7f2a99f944df64f6af4bf634c5bf4ad8f6782998cd9e860a8f7ef005ecf9
  • Pointer size: 130 Bytes
  • Size of remote file: 10.2 kB
app/static/team/katsanevakis.jpg ADDED

Git LFS Details

  • SHA256: 6d7d52365672e877e3cc2ccc2ec8b9e98a2e2b447772352c447e732f9facb08c
  • Pointer size: 130 Bytes
  • Size of remote file: 23.2 kB
app/static/team/kofidis.jpg ADDED

Git LFS Details

  • SHA256: a234b1bb9bc8f52a3f06456c414883b85d46b2537c210f41ddc9e1bef0d6b3f8
  • Pointer size: 131 Bytes
  • Size of remote file: 124 kB
app/static/team/mazaris.jpg ADDED

Git LFS Details

  • SHA256: e07ad0e7fb13d4845b45a5dfcd80e36a1d71b33d070c9ce052f7d4ffacbb7c72
  • Pointer size: 129 Bytes
  • Size of remote file: 5.22 kB
app/static/team/pantazis.jpg ADDED

Git LFS Details

  • SHA256: e258ce1862ed30bf2d805021e32e0b9633c311b4dd8c3a225955a01ff69c017d
  • Pointer size: 130 Bytes
  • Size of remote file: 63.3 kB
app/static/team/poursanidis.jpg ADDED

Git LFS Details

  • SHA256: 6f37dc6aa58aaee44522aa9d5381243ca00f3dda540954057dabf0d7de893757
  • Pointer size: 130 Bytes
  • Size of remote file: 13.6 kB
app/static/team/zotou.jpg ADDED

Git LFS Details

  • SHA256: 4d52e23f90d9729cadf4d4bb94d634426b32a6a16ab6740acfd59e9552551364
  • Pointer size: 129 Bytes
  • Size of remote file: 8.83 kB
requirements.txt CHANGED
@@ -16,3 +16,6 @@ huggingface-hub
16
  peft
17
 
18
  python-multipart
 
 
 
 
16
  peft
17
 
18
  python-multipart
19
+ ultralytics
20
+ opencv-python-headless
21
+ pyyaml