MetaReview:

I concur with Rev 1 summary that captures the consensual view. "the technical innovations of this paper are not huge, but the contributions are substantially useful in practice and during deployment. Specifically, the complexity-based exit strategy as well as the computational savings achieved by MOOD will be useful to the CV community working on OOD detection." I also appreciate the quality of the rebuttal document that has help addressing the queries raised by the reviewers in terms of presentation, reproducibility of the results, and committment to share the code.