Lue Fan
Lue Fan
Simply sample the 0, 2, 4, ... , 2n columns of coord_s1.
I have no experience with torchscript. However, it is a very simple op, so it should be easy to implement it in your way. @rockywind
Our code has done the job, we crop the front view here: https://github.com/tusen-ai/RangeDet/blob/main/datasets/create_range_image_in_kitti.py#L156
It is designed to create a fake foreground point at the start position of each sample. So, if the inference batch_size is 1, they are the same.
@WYRTDCQ Please follow the instructions first: https://github.com/tusen-ai/SST/blob/main/docs/overall_instructions.md#basic-usage Specifically, you should install TorchEx.
FSD on Argoverse2 val split: ``` AP ATE ASE AOE CDS ARTICULATED_BUS 0.204 0.765 0.201 0.277 0.159 BICYCLE 0.386 0.270 0.237 0.461 0.320 BICYCLIST 0.334 0.233 0.254 0.367 0.280 BOLLARD...
You only need to re-generate `waymo_dbinfo_train.pkl`. If you know well about the format, you only need to modify the coordinates in this pickle file instead of regeneration. FYI, here are...
You could try an experiment without dbsampler. I doubt there is something wrong with it.
Send me an email and I will offer you the trained checkpoint. You could use it for inference to see if the results match.
I have sent it to you.