NeurIPS 2020

Deep Variational Instance Segmentation

Meta Review

The proposed formulation is novel and interesting. The empirical evaluation shows promising performances compared to other one-stage IS methods. However, the reviewers agree that it is closely related to proposal-based approaches and additional comparisons would make the paper's conclusion more convincing. Still, in the discussion the reviewers come to the conclusion that the paper provides a valuable contribution that is marginally above the acceptance threshold.