search

andyzeng / visual-pushing-grasping

Train robotic agents to learn to plan pushing and grasping actions for manipulation with deep reinforcement learning.
http://vpg.cs.princeton.edu/
BSD 2-Clause "Simplified" License
898 stars 315 forks source link

image size #27

Open mike112223 opened 5 years ago

mike112223 commented 5 years ago

Hi Andy! Could you please give some insights about doubling the input size? Zoom image by 2, go through network and then upsample by 16 to recover the resolution. Why simply do network forward and upsample by 32?

mabouseif commented 4 years ago

@mike112223 I am not sure, but maybe because it's less interpolation when you upsample by 16 rather than by 32, so you do not "assume" alot of information.