search

Megvii-BaseDetection / BEVDepth

Official code for BEVDepth.
MIT License
688 stars 96 forks source link

can I just use keyframes? Can only keyframes be detected and constraints on depth?Thx #126

Open emilyemliyM opened 1 year ago

emilyemliyM commented 1 year ago

What is the purpose of using non-keyframes and can I just use keyframes? Can only keyframes be detected and constraints on depth be applied?

Thanks.

yinchimaoliang commented 1 year ago

Hi there, do you mean using only one frame?

emilyemliyM commented 1 year ago

Hi there, do you mean using only one frame?

yes Using the key frame only without the non-key frames.

If so, it means that to set "max_cam_sweeps=0, max_lidar_sweeps=0" in the "gen_info.py"?

yinchimaoliang commented 1 year ago

There is no need to change gen_info.py. You can do this by simply changing here https://github.com/Megvii-BaseDetection/BEVDepth/blob/main/bevdepth/exps/nuscenes/mv/bev_stereo_lss_r50_256x704_128x128_24e_key4.py#L33-L35.

emilyemliyM commented 1 year ago

There is no need to change gen_info.py. You can do this by simply changing here https://github.com/Megvii-BaseDetection/BEVDepth/blob/main/bevdepth/exps/nuscenes/mv/bev_stereo_lss_r50_256x704_128x128_24e_key4.py#L33-L35.

Thank you very much for your prompt reply.

Last question, what is the purpose of using non-keyframes? I saw a sentence in the code comments and I'm a little confused. image

Does it mean that there is a loss of a specific-view camera data in keyframes? maybe "cam_back_left" ?

yinchimaoliang commented 1 year ago

The non-keyframes are frames that don't have annotations. The code above is used to handle scenarios that some frames don't have all six cameras' data, so we have to remove these frames.