BGNet: Optimized for Mobile Deployment
Segment images in real-time on device
BGNet or Boundary-Guided Network, is a model designed for camouflaged object detection. It leverages edge semantics to enhance the representation learning process, making it more effective at identifying objects that blend into their surroundings
This model is an implementation of BGNet found here.
This repository provides scripts to run BGNet on Qualcomm® devices. More details on model performance across various devices, can be found here.
WARNING: The model assets are not readily available for download due to licensing restrictions.
Model Details
- Model Type: Model_use_case.semantic_segmentation
- Model Stats:
- Model checkpoint: BGNet
- Input resolution: 416x416
- Number of parameters: 77.8M
- Model size (float): 297 MB
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |
|---|---|---|---|---|---|---|---|---|
| BGNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 117.969 ms | 1 - 143 MB | NPU | -- |
| BGNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 114.959 ms | 2 - 78 MB | NPU | -- |
| BGNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 33.59 ms | 1 - 218 MB | NPU | -- |
| BGNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 38.675 ms | 2 - 68 MB | NPU | -- |
| BGNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 23.009 ms | 1 - 23 MB | NPU | -- |
| BGNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 20.17 ms | 2 - 28 MB | NPU | -- |
| BGNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 18.796 ms | 0 - 180 MB | NPU | -- |
| BGNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 154.396 ms | 1 - 142 MB | NPU | -- |
| BGNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 32.303 ms | 1 - 79 MB | NPU | -- |
| BGNet | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 117.969 ms | 1 - 143 MB | NPU | -- |
| BGNet | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 114.959 ms | 2 - 78 MB | NPU | -- |
| BGNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 23.045 ms | 1 - 23 MB | NPU | -- |
| BGNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 20.194 ms | 2 - 28 MB | NPU | -- |
| BGNet | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 37.507 ms | 0 - 110 MB | NPU | -- |
| BGNet | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 33.987 ms | 2 - 55 MB | NPU | -- |
| BGNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 22.985 ms | 0 - 23 MB | NPU | -- |
| BGNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 20.31 ms | 2 - 28 MB | NPU | -- |
| BGNet | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 154.396 ms | 1 - 142 MB | NPU | -- |
| BGNet | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 32.303 ms | 1 - 79 MB | NPU | -- |
| BGNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 16.668 ms | 1 - 227 MB | NPU | -- |
| BGNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 14.497 ms | 2 - 94 MB | NPU | -- |
| BGNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 13.941 ms | 3 - 98 MB | NPU | -- |
| BGNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 12.899 ms | 0 - 140 MB | NPU | -- |
| BGNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 11.566 ms | 2 - 84 MB | NPU | -- |
| BGNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 11.082 ms | 2 - 79 MB | NPU | -- |
| BGNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 11.08 ms | 1 - 141 MB | NPU | -- |
| BGNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 8.872 ms | 2 - 89 MB | NPU | -- |
| BGNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 9.291 ms | 2 - 86 MB | NPU | -- |
| BGNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 20.719 ms | 403 - 403 MB | NPU | -- |
| BGNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 19.194 ms | 154 - 154 MB | NPU | -- |
Installation
Install the package via pip:
# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
pip install pysodmetrics==1.5.1 --no-deps
pip install "qai-hub-models[bgnet]"
Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
Sign-in to Qualcomm® AI Hub Workbench with your
Qualcomm® ID. Once signed in navigate to Account -> Settings -> API Token.
With this API token, you can configure your client to run models on the cloud hosted devices.
qai-hub configure --api_token API_TOKEN
Navigate to docs for more information.
Demo off target
The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input.
python -m qai_hub_models.models.bgnet.demo
The above demo runs a reference implementation of pre-processing, model inference, and post processing.
NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).
%run -m qai_hub_models.models.bgnet.demo
Run model on a cloud-hosted device
In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following:
- Performance check on-device on a cloud-hosted device
- Downloads compiled assets that can be deployed on-device for Android.
- Accuracy check between PyTorch and on-device outputs.
python -m qai_hub_models.models.bgnet.export
How does this work?
This export script leverages Qualcomm® AI Hub to optimize, validate, and deploy this model on-device. Lets go through each step below in detail:
Step 1: Compile model for on-device deployment
To compile a PyTorch model for on-device deployment, we first trace the model
in memory using the jit.trace and then call the submit_compile_job API.
import torch
import qai_hub as hub
from qai_hub_models.models.bgnet import Model
# Load the model
torch_model = Model.from_pretrained()
# Device
device = hub.Device("Samsung Galaxy S25")
# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()
pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
# Compile model on a specific device
compile_job = hub.submit_compile_job(
model=pt_model,
device=device,
input_specs=torch_model.get_input_spec(),
)
# Get target model to run on-device
target_model = compile_job.get_target_model()
Step 2: Performance profiling on cloud-hosted device
After compiling models from step 1. Models can be profiled model on-device using the
target_model. Note that this scripts runs the model on a device automatically
provisioned in the cloud. Once the job is submitted, you can navigate to a
provided job URL to view a variety of on-device performance metrics.
profile_job = hub.submit_profile_job(
model=target_model,
device=device,
)
Step 3: Verify on-device accuracy
To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device.
input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
model=target_model,
device=device,
inputs=input_data,
)
on_device_output = inference_job.download_output_data()
With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output.
Note: This on-device profiling and inference requires access to Qualcomm® AI Hub Workbench. Sign up for access.
Run demo on a cloud-hosted device
You can also run the demo on-device.
python -m qai_hub_models.models.bgnet.demo --eval-mode on-device
NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).
%run -m qai_hub_models.models.bgnet.demo -- --eval-mode on-device
Deploying compiled model to Android
The models can be deployed using multiple runtimes:
TensorFlow Lite (
.tfliteexport): This tutorial provides a guide to deploy the .tflite model in an Android application.QNN (
.soexport ): This sample app provides instructions on how to use the.soshared library in an Android application.
View on Qualcomm® AI Hub
Get more details on BGNet's performance across various devices here. Explore all available models on Qualcomm® AI Hub
License
- The license for the original implementation of BGNet can be found [here](This model's original implementation does not provide a LICENSE.).
- The license for the compiled assets for on-device deployment can be found here
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.