search

YuemingJin / STswinCL

[TMI'22]Exploring Intra- and Inter-Video Relation for Surgical Semantic Scene Segmentation
21 stars 2 forks source link

A question regarding the data setting #3

Open BeileiCui opened 1 year ago

BeileiCui commented 1 year ago

Hi. In seg18/dataset/Endovis2018_new.py the 15 training videos are seperated into 12 train set and 3 validation set. However, during training the network were trained with all the 15 train set and make the validation on the 4 test set. May I ask the reason for doing this?

YuemingJin commented 1 year ago

Hi, thanks for your interest in our work. In the real training procedure, we exactly follow the evaluation protocol of the Challenge - train by 15 videos, and test on 4 videos. We combine 12/3 to 15 train set in Line 51. You could ignore this separation, which is just used for our initial attempt. Sorry for the confusion.