Get part of detected human body

Hallo everyone, i am new here and new on opencv also.
i need a model to detect shoe on visitor/people via CCTV video and then get the color of shoe, but i don’t find the match model for this case.

So, i have an idea, i will detect human/people with existing model (on internet) then i will crop lower part of detected people and i assumed it as shoe.

Is it possible? if yes, should i use additional library or opencv is enough

Thank you…

It seems possible. You can start with opencv_zoo/models/pose_estimation_mediapipe at main · opencv/opencv_zoo · GitHub

thanks for your reply laurent,
that is a cool models,

i will try and be back after this technique work properly

i have already tried this library, but it take a long time processing and make the video freeze.
because for each frame i did.

  1. get landmark 27 until 32
  2. get bounding box and crop image
  3. get color from that image

is any wrong from my logic? or any other solution, model or library?


try GitHub - CMU-Perceptual-Computing-Lab/openpose: OpenPose: Real-time multi-person keypoint detection library for body, face, hands, and foot estimation