MetaReview:

The authors present a method for out-of-distribution detection. The paper is well-written, well-explained and well-illustrated and the idea is novel, simple and well-motivated. The experimental setup is sufficient and comprehensive, the analysing and fundings are interesting and the results are convincing. The method is especially suitable for a large number of semantic classes. After considering the rebuttal and other reviews, all reviewers unanimously recommend acceptance. The ACs agree with the reviewers and think that it would be beneficial if the authors could include the additional ablations, detailed results and references of the rebuttal in the camera-ready version and release the code and datasets as promised.