LibreYOLO live demo

Pick a model, upload an image, and run inference. Covers the two flagship families plus depth:

  • YOLOv9 (CNN detector)
  • RF-DETR (transformer detector) for detection, instance segmentation and keypoints
  • Depth Anything V2 for monocular depth

Weights download automatically from huggingface.co/LibreYOLO the first time you pick a model, so the first run of each model is slower. This Space runs on CPU, so the RF-DETR keypoints model (size x) can take a while per image.

Model
0.05 0.95

LibreYOLO

Website · GitHub · Hugging Face · Benchmarks (Vision Analysis) · LinkedIn · Reddit

MIT-licensed open-source computer vision. If you find it useful, a star on GitHub helps a lot.