Vsion FSD

why train on a subset of video(id 22-id42 in the breakdance-flare)? why not train with the whole video

Does the model only handle the situation when the camera is not moving?

batchsize 4 only cost 4Ggpu memory. you can improve the batchsize. Are you saying that the larger the batchsize, the better performance?

I used your preprocess code (https://github.com/gengshan-y/viser-release/blob/main/preprocess/README.md) but got stuck (pdb) where you break. Commenting out can be trained, but it seems that the program has bugs. you can try it...

Isn't this place data preprocessing code? (https://github.com/gengshan-y/viser-release/blob/main/preprocess/README.md)

> Re Isn't this place data preprocessing code? (https://github.com/gengshan-y/viser-release/blob/main/preprocess/README.md) I need to make it clear that the codebase's preprocessing code is complete

the result on car-turn is far from useful

at the beginning, the train/flowobs map is black (on tensorboard).