9 ms·
There are different "problems" as you're saying that could tackle your needs. - Gesture recognition to detect different hand gestures - Skeleton detection to
by nonoesp 2y ago
There are different "problems" as you're saying that could tackle your needs.
- Gesture recognition to detect different hand gestures
- Skeleton detection to infer people's positions and whether they have a raised hand
- Object detection (e.g., YOLO) to detect hands
I believe in this case you could go for skeleton detection as you get information from which you could infer raised hands.
- swyx 2y agoright but can skeleton detection do hundreds of people at once?
- SkalskiP 2y agoYou can always slice the images into smaller ones, run detection on each tile, and combine results. Supervision has a utility for this - https://supervision.roboflow.com/latest/detection/tools/inference_slicer/ https://supervision.roboflow.com/latest/detection/tools/infe..., but it only works with detections. You can get a much more accurate result this way. Here is some side-by-side comparison: https://github.com/roboflow/supervision/releases/tag/0.14.0 https://github.com/roboflow/supervision/releases/tag/0.14.0.