Reproduce Akanazawa Motion Reconstruction: Day 1

Original Project: SFV: Reinforcement Learning of Physical Skills from Videos

Original Project Code: https://github.com/akanazawa/motion_reconstruction

Issues

• Original Project Code written for Python v1
• Original Project Code written for Python 2 (allows mixed use of tabs and spaces)
• Python 3 too many missing dependencies
• Could not find video files
• Could not find neutral_smpl_with_cocoplus_reg.pkl in smpl package

Solutions

• Install all dependencies using pip3 or sudo apt-get
• Replaced all tabs with spaces
• Found videos here and renamed them to fit code
• Changed:

-import tensorflow as tf
+import tensorflow.compat.v1 as tf

-kVidDir = '/home/kanazawa/projects/hmr_sfv/demo_data/videos'
-kOutDir = '/home/kanazawa/projects/hmr_sfv/demo_data/openpose_output'
+kVidDir = './demo_data/videos'
+kOutDir = './demo_data/openpose_output'

-       overlap = (w * h) / area[idxs[:last]]
+        overlap = (w * h) / area[idxs[:last]]

-       idxs = np.delete(idxs, np.concatenate(([last],
+        idxs = np.delete(idxs, np.concatenate(([last],

Progress

• Reached video file processing in Motion Reconstruction Project

Next Steps

• Move on to OpenPose stage of this project