Corentin Sautier

For future reference, the results of scipy's LSA and of this repo's output have different formatting (see https://github.com/gatagat/lap/issues/10 for reference). You can correct your code as follows, and you'll see...

Replace `if pkl_file.exists():` by if `os.path.exists(pkl_file)` Later on you might also have to comment the line `info_path = self.check_sequence_name_with_all_version(info_path)` and replace `sequence_file_tfrecord = sequence_file[:-9] + '_with_camera_labels.tfrecord` by `sequence_file_tfrecord = sequence_file[:-9]...

Yes, I've found too that the waymo preprocessing cost a lot of memory. A partial solution would be to remove the multiprocessing by replacing ``` with futures.ThreadPoolExecutor(num_workers) as executor: sequence_infos...

As for using the OpenPCDet preprocessing I have no idea. I'm not affiliated with the authors of the code, I'm just trying to get the code running as well.

I can't tell for sure, but the `waymo_one_sweep_mm.yaml` seems to use a bit less than 3T. Maybe start with KITTI as it seems to be much lighter and probably easier...

The pre-processing last about 150 hours on my hardware, with no multiprocessing. I'm not sure why it uses any GPU memory, as far as I can tell the pre-processing is...

Hi, thank you for your interest in our work. If union is zero, it means that there is no element of this class in the ground truth, and we predicted...

This is intended, as the 0 row of the confusion matrix correspond to GT ignore points which shouldn't influence the score. However the 0 column of the matrix correspond to...