This repository showcases the robustness of Openpose 2D human pose estimator against limb occlusions. Adding 20*16 rectangle mask at center of both thighs only reduce the detection rate by 0.105.
Description
Run the script pose_estimator_test.py and sees everything for yourself.
The required pose-estimator model file can be downloaded at: https://yadi.sk/d/blgmGpDi3PjXvK
Prerequsites
tqdm
Keras
numpy
skimage
h5py
Please note that Keras is merged into tensorflow now, so it's now from tensorflow.keras instead of from keras