Does the code have corresponding paper?
liu09114 opened this issue · 6 comments
Or do you have a result like room layout image?
It seems to be a part of IM2CAD(chapter3.1 in the paper. Use LSUN dataset and Fully Convolutional Networks to label each pixel). Or maybe it's just a coincidence.
@liu09114 We've updated the project though, and it's still in progress.
@yyong119 Nope, not really. There have been several FCN-like approaches for layout estimation. And IM2CAD use roughly the same idea in DeLay (Dasgupta et al.) for this part. Our design is inspired more by ST-PIO (Zhao et al.).
May refer,
http://lsun.cs.princeton.edu/leaderboard/index_2016.html#roomlayout
http://lsun.cs.princeton.edu/leaderboard/index.html#roomlayout
@leVirve Thank you. I found that few ground truth label mistakes in the LSUN datasets, did you correct them?
Yes, there are some mislabeled layouts in original LSUN Room, and we correct them manually.
could you please share the labeled LSUN dataSet about the room , I 'll be very very thankful to your help!!!!! And my email adress is : t.f-zhan@foxmail.com