Hi, @xvjiarui , I have read the paper and I am curious about the evaluation for the VIP dataset, could you share the evaluation code for this dataset?

Something is not clear here, I hope you can help to answer my following concerns.
In your paper, not the only the first gt mask but also the prediction results in the latest 20 frames are propagated to the current frame. It seems that previous methods only use the first gt mask with several (1 in Joint-task Self-supervised Learning for Temporal Correspondence ) prediction results to do the propagation, is there any ablation study about this number?

Pretrained model for the VIP dataset #14

Description

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions