zhanghang1989 / PyTorch-Encoding

A CV toolkit for my papers.
https://hangzhang.org/PyTorch-Encoding/
MIT License
2.04k stars 452 forks source link

pascol voc aug dataset #220

Open alphaccw opened 5 years ago

alphaccw commented 5 years ago

from the paper, pascal aug contains 10582, 1449, 1456 train, val, test images. But when I check the code, it seems use 11355 (another split ) for training. Did I miss something?

BebDong commented 4 years ago

There are 8498 images for training and 2857 images for validation in SBD dataset, which is 11355 in total. But some images occur in both SBD and Pascal VOC 2012, resulting in 10582 images for training practically.