search

Banconxuan / RTM3D

The official PyTorch Implementation of RTM3D and KM3D for Monocular 3D Object Detection
MIT License
454 stars 85 forks source link

Something missing and how to convert kps into positions #16

Open mrsempress opened 3 years ago

mrsempress commented 3 years ago

Thanks for your excellent work. Can you give me some suggestions on how to convert kps to positions? Either relationship diagram or tutorial website is both Ok. I would appreciate if you would like.

Maybe I missed some knowledge, but I think there are some problems in the code.

  1. In trains/car_pose.py line129, 79, you called car_pose_decode() function, but the definition in models/decode.py is like that car_pose_decode( heat, wh, kps, dim, rot, prob=None, reg=None, hm_hp=None, hp_offset=None, K=100, meta=None, const=None), You didn’t pass the parameter rot.
  2. And in models/decode.py line 590, you define as gen_position(kps, dim, rot, meta, const), the const use the default ‘None’, but 'NoneType' object has no attribute 'expand’.
vobecant commented 3 years ago

I have the same problem.

lucasjinreal commented 3 years ago

I think this code is not runable. @Banconxuan Please make sure your code can be run when opensource..

kai4729 commented 3 years ago

I meet it too, anybody solve it?