search

charigyang / motiongrouping

Repository accompanying the paper: Self-supervised Video Object Segmentation by Motion Grouping. C. Yang, H. Lamdouar, E. Lu, A. Zisserman, W. Xie.
BSD 2-Clause "Simplified" License
238 stars 34 forks source link

Training on all DAVIS sequences? #11

Open danielbear opened 2 years ago

danielbear commented 2 years ago

Hello,

Thank you for sharing the code for this exciting work! I'm hoping to use your model as input to another model but want to make sure I can reproduce your results on DAVIS first.

As I understand it, the DAVIS trn_loader trains on all the sequences in DAVIS, not just the ones specified in DAVIS/ImageSets/1080p/train.txt. This will include the validation sequences as well. Am I understanding this correctly, or do you actually hold out the validation sequences during training (and then report DAVIS results on those?)

Best, Dan

YudeWang commented 2 years ago

Same question. I wonder whether it is the common setting for unsupervised video object segmentation.