NeurIPS 2020

Labelling unlabelled videos from scratch with multi-modal self-supervision

Meta Review

The initial scores for this paper were: 6: Marginally above the acceptance threshold. 7: A good submission; accept. 7: A good submission; accept. 7: A good submission; accept. Reviewers are overall positive but raise several issues: R1 raises a number of points in the description of the method that need to be clarified, regarding the description of the approach, empirical evaluation, practicability of the approach, and significance of the contribution. R2 suggests additional ablations / baselines. R3 suggests emphasizing better the contribution of the work and has questions regarding the description of the approach. R4 suggests moderating claims of the work and asks for clarifying the relation w.r.t. some prior work. The authors provide a rebuttal, which addresses many of these concerns. The reviewers keep their positive ratings. R1 still finds that one issue was not completely addressed in the rebuttal and maintains their borderline positive score (6). The AC is overall convinced by the positive arguments of the reviewers and recommends Accept. The authors are encouraged to take into account all reviewers' feedback (and especially the remaining issue mentioned by R1) when preparing the final version.