OOP'ification from Zhe Cao's code using the model from Michal Faber. Please respect the initial authors licenses. Install as follows:
- Python 3.6
- Keras 2
pip install git+https://github.com/jutanke/easy_multi_person_pose_estimation
import cv2
from poseestimation import model
img_path = './julian.JPG'
I = cv2.cvtColor(cv2.imread(img_path), cv2.COLOR_BGR2RGB)
# I can either be a single image or a list of images
pe = model.PoseEstimator()
# end-to-end prediction of all detected poses
positions = pe.predict(I)
# extract the heatmaps and part affinity fields
heatmaps, pafs = pe.predict_pafs_and_heatmaps(I)