LibreRFDETRx-pose

RF-DETR keypoint preview model for COCO 17-keypoint person pose estimation, repackaged for LibreYOLO.

This is an early-access PREVIEW model, matching the upstream release. Its behavior and interface may change without compatibility guarantees.

Source

Derived from roboflow/rf-detr at v1.8.0, Copyright (c) 2024-2025 Roboflow Inc., licensed under the Apache License, Version 2.0.

The RF-DETR architecture uses a DINOv2 vision backbone, Copyright (c) Meta AI, also licensed under the Apache License, Version 2.0.

Modifications

The original RF-DETR keypoint state dict is repackaged into the LibreYOLO checkpoint format with LibreYOLO metadata (model family, size, task, class names, and keypoint configuration). The underlying weight values are unchanged. See weights/convert_rfdetr_keypoint_weights.py in the LibreYOLO repository for the exact repackaging procedure.

License

This model is distributed under the Apache License, Version 2.0. See the LICENSE and NOTICE files for the full license text and attribution.