How to prepare the data?

arthurdouillard / CVPR2021_PLOP

Official code of CVPR 2021's PLOP: Learning without Forgetting for Continual Semantic Segmentation

https://arxiv.org/abs/2011.11390

MIT License

145 stars 23 forks source link

How to prepare the data? #2

Closed zhangchbin closed 3 years ago

zhangchbin commented 3 years ago

I noticed that there are some differences in data. How can I prepare the data?

arthurdouillard commented 3 years ago

You don't really have to prepare anything per se. For VOC and ADE you can use Cermelli's script which are in the data folder. For Cityscapes I've simply download the link they gave me (you need to send some request to Cityscapes org), and unzip all compressed files.

Then you need to specify in command line, which dataset you want and what is the DATA_ROOT.

zhangchbin commented 3 years ago

Hi, this is my 'data' folder . DATA_ROOT is 'data/'

arthurdouillard commented 3 years ago

Your DATA_ROOT for, for example VOC, should contain the folders SegmentationClassAug, VOCdevkit, and list.

The 1st contain segmentation masks, the second the pixel images, and the third the train/val split.

zhangchbin commented 3 years ago

Thank you very much! I have solved this problem!

arthurdouillard commented 3 years ago

Great! I'll to add more explanation in the readme if I can make it clearer.

Don't hesitate to open a new issue, if you have other problems :)