search

wenbowen123 / icra20-hand-object-pose

[ICRA 2020] Robust, Occlusion-aware Pose Estimation for Objects Grasped by Adaptive Hands
Other
52 stars 6 forks source link

i think here is a bug. #5

Open jih189 opened 3 years ago

jih189 commented 3 years ago

https://github.com/wenbowen123/icra20-hand-object-pose/blob/da59809d862ee7232eb2c3b8751f30acebf87a90/src/perception/src/Hand.cpp#L749

float pitch = rpy(1); // Rotation along y axis pitch = std::min(std::abs(pitch), std::abs(static_cast(M_PI)-pitch)); pitch = std::min(std::abs(pitch), std::abs(static_cast(M_PI)+pitch));

if pitch is -3.14036 at the beginning, then the pitch will be 3.14036. I think it should be 0.00036 instead.

Please check it out

jih189 commented 3 years ago

hi, wenbowen. Do you really use this code to get the result in your paper? I found some other bug in your code, and I wonder how you got your result. Please fix the function rejectedbyrender()

wenbowen123 commented 3 years ago

Hi sorry for the late reply. It has been quite some time since I last time worked on this project. It looks like you are right, but I'll need to take a closer look and get back to you. That said, the entire framework is sort of robust due to the multi-hypothesis generation and pruning. So even if some hypothesis are falsely kept, they'll be further evaluated based on LCP score. However, fixing this would hopefully make some performance gain.