search

JiayuZou2020 / HFT

[ICRA 2023] Official Pytorch implementation for HFT
https://arxiv.org/abs/2204.05068
MIT License
123 stars 13 forks source link

About image resize with intrinsics in config. #4

Open QixingJiang opened 1 year ago

QixingJiang commented 1 year ago

Thank you for your contribution to the community through your work! But I have some doubts about the code in config: HFT/configs/pyva/pyva_swin_kd_simple_fpn_force_nuscenes.py https://github.com/JiayuZou2020/HFT/blob/c4d06714e302e9f659635d0254e0dcf63089c257/HFT/configs/pyva/pyva_swin_kd_simple_fpn_force_nuscenes.py#L49-L50 These configurations resize the images of nuScenes to 1024 and in the load function: https://github.com/JiayuZou2020/HFT/blob/c4d06714e302e9f659635d0254e0dcf63089c257/HFT/mmseg/datasets/pipelines/loading.py#L202-L207 Here, the calib (intrinsics) is multiplied by a scaling factor, provided that the image is resized to a fixed size of 800600, not 10241024. This hardcore caused some misalignment, did I miss something? Thank you for answering my question.

QixingJiang commented 1 year ago

In fact, when the resize scale is 1024 1024, the camera parameters scaled to incorrect 600 800 in the current code will be better, and I don't know why ... Can you give me some clues?