--- license: apache-2.0 library_name: libreyolo pipeline_tag: object-detection tags: - object-detection - deformable-detr - detr datasets: - detection-datasets/coco --- # LibreDeformableDETRr50ssdc5 The official single-scale DC5 ResNet-50 Deformable DETR checkpoint (41.5 AP on COCO val2017), converted for [LibreYOLO](https://github.com/LibreYOLO/libreyolo). ```python from libreyolo import LibreYOLO model = LibreYOLO("LibreDeformableDETRr50ssdc5.pt") results = model.predict("image.jpg") ``` LibreYOLO ships this museum family for inference only. Its portable deformable attention uses the upstream pure-PyTorch `grid_sample` reference path; the custom CUDA extension is not packaged. Native and ONNX inference use a fixed 800 x 800 PIL-bilinear stretch with ImageNet normalization. That deployment transform intentionally differs from upstream's short-side-800/max-1333 evaluation transform. ## Source Architecture derived from [fundamentalvision/Deformable-DETR](https://github.com/fundamentalvision/Deformable-DETR) at commit `11169a60c33333af00a4849f1808023eba96a931`. Copyright (c) 2020 SenseTime. All Rights Reserved. Modified from DETR, Copyright 2020 - present, Facebook, Inc. Licensed under Apache License 2.0. The checkpoint comes from [SenseTime/deformable-detr-single-scale-dc5](https://huggingface.co/SenseTime/deformable-detr-single-scale-dc5) at revision `c23332913d0ae1a8c98725e308eccba65a5933cc`. Source `model.safetensors` SHA-256: `e71afa5f5900e2e769275156494195508efcadaab4275b0cd4c80f10369dc090`. ## Modifications Checkpoint keys are remapped to the original module layout, decoder Q/K/V projections are concatenated, and serialization aliases for shared prediction heads are restored. Non-parameter BatchNorm tracking counters are omitted; learned tensor values are unchanged. The released 91-column COCO category-id head is retained and mapped to contiguous COCO-80 classes during postprocess. The converted checkpoint SHA-256 is `777237ddcb4cfe84d72a1daad5e447cd1a070b212def10102981779c405eb5bf`. LibreYOLO's output tensors are bit-exact against the pinned upstream pure-PyTorch path (`max_abs_diff == 0.0`); fixed-800 ONNX Runtime prediction parity is also verified. See `weights/convert_deformable_detr_weights.py` and `docs/provenance/deformable_detr.md` in the [LibreYOLO source repository](https://github.com/LibreYOLO/libreyolo). ## License Apache License 2.0. See the [`LICENSE`](./LICENSE) and [`NOTICE`](./NOTICE) files in this repository.