to_coco error, numpy binary incompatibility about bdd100k

[fyu]

Upgrading numpy to version 1.20.3 appears to have solved the problem. Perhaps pycocotools has been updated with a newer version of numpy?

Yes, pycocotools looks like the culprit.

[fyu]

This is because your install python libraries are depending on different numpy versions. You can resolve this issue by creating a clean virtual python environment and install all the required libraries with pip install -r requirements.txt from scratch.

[pcicales]

Hello @fyu , thank you for the wonderful dataset and rapid reply!

That is what I did using an anaconda env, still using the same command you listed. Here is the version info:

Python 3.7.10 (default, Feb 26 2021, 18:47:35)
[GCC 7.3.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
import numpy
print(numpy._ _ version _ _)
1.19.5

[fyu]

You have the right numpy version, but some of your libraries (maybe pycocotools) depend on a different version of numpy. A clean install usually solves the problem.

[pcicales]

Upgrading numpy to version 1.20.3 appears to have solved the problem. Perhaps pycocotools has been updated with a newer version of numpy? Here is the output of the same corrected command:

(bdd100k) redact@redact:/redact/workspace/pycharm_projects/bdd100k$ python3 -m bdd100k.label.to_coco -i /redact/bdd100k/labels/det_20/det_train.json -o /redact/datasets/bdd100k/labels/det_20_coco/det_train.json [2021-05-31 14:33:39,024 to_coco.py:577 main] Loading annotations... [2021-05-31 14:34:37,101 to_coco.py:586 main] Start format converting... 100%|██████████████████████████████████████████████████████████████████████████| 69863/69863 [00:01<00:00, 53779.44it/s] 100%|███████████████████████████████████████████████████████████████████████████| 69863/69863 [00:10<00:00, 6383.82it/s] [2021-05-31 14:34:49,386 to_coco.py:590 main] Saving converted annotations to disk... [2021-05-31 14:35:12,709 to_coco.py:593 main] Finished!