The efficient way to get the estimated `.label` file about segregator HOT 5 CLOSED

Hi HyungTae!

Thanks for reaching out!
Of course I know you haha :) I noticed your work a long time ago and Quatro is really nice work!
For the semantic label generation, I did not use Salsanext but SPVNAS (on KITTI), where I did not encounter the number inconsistency issue. If you need any help regarding SPVNAS, do not hesitate to discuss it here!

from segregator.

Thx for prompt reply and it works!

Here's my code snippet in evaluate.py in SPVNAS repository for other researchers:

abs_input_path = feed_dict['file_name'][0].split("/")
output_file_name = args.save_dir + "/" + abs_input_path[-3] + "/" + abs_input_path[-1].replace('.bin', '.label')
sem = outputs.cpu().numpy().astype(np.uint32)
ins = np.zeros_like(sem).astype(np.uint32)
pred_eval = sem + (ins << 16)
pred_eval.astype(np.uint32).tofile(output_file_name)

• I added args.save_dir parameter
• abs_input_path[-3]: sequence number, e.g., 00, 01, ...
• abs_input_path[-1]: frame number, e.g., 000000, 000001, ..., 004000, ...

from segregator.

Like this:

from segregator.

This issue ganna be closed; yet, it would be better if you provide one with pre-extracted labels in the future :)

from segregator.

Hi HyungTae!

Thanks for your advice. And I put this issue on the top to let more people know :)

from segregator.