AnnotateIt/rtdetr-v2-r50vd-coco-onnx
<!-- annotateit-brand:start --> <p><a href="https://huggingface.co/AnnotateIt"><img src="https://huggingface.co/spaces/AnnotateIt/README/resolve/d399a89b9d5dcf41afbe7cf29bf99c9028bc51e2/assets/annotateit-logo.png" width="48" height="48" alt="AnnotateIt"></a></p>
[AnnotateIt](https://annotateit.ai/) · [Open the app](https://app.annotateit.ai/) · [Models & datasets](https://huggingface.co/AnnotateIt) · [Documentation](https://annotateit.ai/docs/) <!-- annotateit-brand:end -->
RT-DETRv2 r50vd COCO — verified ONNX mirror for AnnotateIt
Byte-identical mirror of the FP32 ONNX graph from `onnx-community/rtdetr_v2_r50vd-ONNX` (revision 1699ea019e5570bb7283d80ee9a9e42d6245436f), based on the official Apache-2.0 `PekingU/rtdetr_v2_r50vd` checkpoint. AnnotateIt did not alter or quantize the graph; this mirror adds an explicit license, immutable provenance, checksums and independent validation. Not an official PekingU / RT-DETR release.
Artifact
Preprocessing and outputs
- Convert to RGB.
- Resize to 640×640, bilinear, stretch (no letterbox).
- Rescale to
[0, 1](÷255); do not apply mean/std normalization. - HWC→NCHW, add batch dim.
sigmoid(logits), best COCO class per query;pred_boxes= normalizedcxcywh→ map to original size.
80 COCO classes, end-to-end set prediction — no NMS.
License
Apache-2.0 (see `LICENSE`); retain attribution when redistributing. Official project: `lyuwenyu/RT-DETR`.
