[Progress-Summary] 16/10

[Lotemn102]

Done

• Analyzed the center of mass data. Why i can't see any difference between olds and youngs?
• Started writing the data reader for this task
• Talked to Alon regarding the PointNet training. He had cool ideas for data augmentations:
  • Add random noise of [-Xmm, Xmm] to each point.
  • Rotate the point cloud in Y axis
  • Apply permutations on each sample. Right now the first row in the each sample is always the same point, and that might hurt the net's ability to generalize.
  • "Construct" new old people, by taking old sample and switching only the points that are "hands" or "legs" with young sample.