NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:7092
Title:Sim2real transfer learning for 3D human pose estimation: motion to the rescue


		
After reviewer discussion and rebuttal this paper received three acceptance recommendations. R1 and R2 are more positive about the paper and acknoweldge the contribution. R3 points out that the impact of using just flow and no person and camera motion is limited. Please consider the post-rebuttal portion of the review to include in a final revision. The method, approach and quality of the paper are high as acknowledged by all reviewers. The only disagreement is on the significance part. The problem of learning from simulated data is relevant and applies to the problem studied. This submission presents a system that makes a step forward in this direction, although from an empirical perspective it may be smaller than anticipated. The empirical results appear conclusive though and all reviewers appreciate the ablation studies that are likely to lead to insights for other works. In summary the reviewers are all on the positive side, although one reviewer being more borderline-positive. The positive aspects of this submission outweigh the negative ones on significance (and slightly novelty), the paper is of high quality and should be accepted.