search

NVlabs / Deep_Object_Pose

Deep Object Pose Estimation (DOPE) – ROS inference (CoRL 2018)
Other
1.03k stars 287 forks source link

How to visualize the layers of DOPE ? #295

Open ArghyaChatterjee opened 1 year ago

ArghyaChatterjee commented 1 year ago

Hi,

I was trying to visualize the layers for beliefmaps and vector field generation part for object keypoint detection. Specially a diagram like this what you did in the CenterPose. Do you have something similar somewhere for DOPE (hand drawn will also work) ?

Screenshot from 2023-03-07 07-55-34

TontonTremblay commented 1 year ago

Yeah there is an option on the inference to have --show_belief this will not show the aff though. I had some code to show also the aff. but I am not sure if it is on the repo. The idea is to do something along these lines: https://stackoverflow.com/questions/28898346/visualize-optical-flow-with-color-model should be somewhat easy. I also had code that would show lines. I am not sure if this is in the repo as well. If you are interested in playing with this I can point to you where you could add this code.