[RNE Rewrite] feat: add keypoint detection task pipeline#1280
Conversation
3eb2f56 to
33e6160
Compare
|
@msluszniak Where should I put the blazeface model, we already have the |
…pport children in ImageViewport
|
Blazeface was never released officially. So I would personally force push this on v0.10.0 tag on Hugging Face. |
Co-authored-by: Mateusz Sluszniak <56299341+msluszniak@users.noreply.github.com>
Could you add which repos are updated? |
|
Fixed. Sorry, I was going to add links after re-exporting models but forgot. Also, I've added the YOLOv8n pose estimation model. When implementing PoC I was sure it was the same as YOLO26, but turns out these are different models, so now we have both. I can remove the YOLOv8n if you want, since YOLO26 is like a better version. |
|
Regarding HF, blazeface is missing content of README.md and just to make sure: Are the reexported models indeed compatible with ET 1.2 or it's typo and it's compatible with 1.3 (currently supported on main)? Also yolo26 has config in xnnpack dir, yolov8 doesn't, same for blazeface. Regarding yolov8, if it's fastee than we can keep it. If not, probably worth removing. |
They were exported using 1.2.0. Also currently the binaries we copy from PoC are also 1.2.0.
Ok, then I will remove it.
Should I add the configs to the ones that don't have it or remove it from the ones that have? Where are these configs used actually? |
|
We should rather add these configs. @NorbertKlockiewicz added recently configs for other models. They help in re-exports, debbugability, are readable for agents. Probably there are some more which Norbert can name. |
Co-authored-by: Mateusz Sluszniak <56299341+msluszniak@users.noreply.github.com>
Rebased onto rne-rewrite incl. #1280 keypoint detection. Rename the poseEstimation FEATURE_MAP entry to keypointDetection (tracks the useKeypointDetector hook) and set backends to xnnpack+coreml+mlx (RF-DETR keypoint ships CoreML + MLX variants) + opencv. Declare it in the computer-vision demo app.
Rebased onto rne-rewrite incl. #1280 keypoint detection. Rename the poseEstimation FEATURE_MAP entry to keypointDetection (tracks the useKeypointDetector hook) and set backends to xnnpack+coreml+mlx (RF-DETR keypoint ships CoreML + MLX variants) + opencv. Declare it in the computer-vision demo app.
Description
Adds kepoint detection task pipeline implementation and corresponding example screen in computer-vision app.
Introduces a breaking change?
Type of change
Tested on
Testing instructions
- https://huggingface.co/software-mansion/react-native-executorch-blazeface
-
https://huggingface.co/software-mansion/react-native-executorch-yolov8n-pose- https://huggingface.co/software-mansion/react-native-executorch-yolo26-pose
Screenshots
Related issues
Closes #1241
Checklist
Additional notes