When will you upload all the code?

[zhangchbin]

No description provided.

[Daniel-xsy]

This repo is working in the process. We need to check the official nuScenes license before we can release the nuScenes-C dataset. You can follow the CREATE.md to generate your own corruption sets (it might take a bit long time). If the dataset is prepared, all the models in ./zoo should be able to run.

[jonyzhang2023]

This repo is working in the process. We need to check the official nuScenes license before we can release the nuScenes-C dataset. You can follow the CREATE.md to generate your own corruption sets (it might take a bit long time). If the dataset is prepared, all the models in ./zoo should be able to run.

looks like you did not upload 'from imagecorruptions' on corruptions.py, is there anyway i can use it? I got a urgent job to use robobev, pls. @Daniel-xsy

[Daniel-xsy]

@jonyzhang2023 You can install it through pip install imagecorruptions as it's a third-party package.

[Daniel-xsy]

Just to provide more information. Our dataset is built with the following configuration:

corruptions = [dict(type='CameraCrash', easy=2, mid=4, hard=5),
               dict(type='FrameLost', easy=2, mid=4, hard=5),
               dict(type='MotionBlur', easy=2, mid=4, hard=5)],
               dict(type='ColorQuant', easy=1, mid=2, hard=3)],
               dict(type='Brightness', easy=2, mid=4, hard=5)],
               dict(type='LowLight', easy=2, mid=3, hard=4)],
               dict(type='Fog', easy=2, mid=4, hard=5)],
               dict(type='Snow', easy=1, mid=2, hard=3)]
]

[jonyzhang2023]

cool!
I will try it. Thanks for your great job on robo-bev.

[Daniel-xsy]

Feel free to reach out and re-open this issue if you have any questions :)