News

Developed a 100% Swift version
https://github.com/otmb/TopDownPoseEstimation

What is this

TopDown Pose Estimation on iOS

• BBox: Yolov7-tiny
• Pose Estimation: ViTPose

Install

$ git clone https://github.com/mbotsu/TopDownPoseExample.git
$ cd TopDownPoseExample/TopDownPoseExample
$ curl -OL https://github.com/mbotsu/KeypointDecoder/releases/download/0.0.1/vitpose-b256x192_fp16.mlmodel
$ curl -OL https://github.com/mbotsu/KeypointDecoder/releases/download/0.0.1/yolov7-tiny_fp16.mlmodel

Example

References

• ViTPose to CoreML
  • mbotsu/20221128_convert.ipynb
• Yolov7 to CoreML
  • john-rocky/CoreML-Models
    • Yolov7 Google Colab

COCO MS val set

mbotsu/KeypointDecoder v0.0.5

Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets= 20 ] = 0.529
Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets= 20 ] = 0.679
Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets= 20 ] = 0.614
Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.479
Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.614
Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 20 ] = 0.593
Average Recall     (AR) @[ IoU=0.50      | area=   all | maxDets= 20 ] = 0.702
Average Recall     (AR) @[ IoU=0.75      | area=   all | maxDets= 20 ] = 0.665
Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = 0.528
Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.684

Model	AP	AP50	AP75	AP(M)	AP(L)
VitPose-b + Yolov7-tiny	52.9	67.9	61.4	47.9	61.4