Learning ODIN

Amir Jevnisek (Tel-Aviv University),* Shai Avidan (Tel Aviv University)
The 33rd British Machine Vision Conference


ODIN is a popular Out-Of-Distribution (OOD) detection algorithm. It is based on the observation that using temperature scaling and adding small perturbations to the input can separate the softmax score distributions between in- and out-of-distribution images, allowing for more effective detection. Instead of passively making this observation, we derive a new loss, termed Gradient Quotient (GQ) loss, that encourages this behaviour by the network. GQ can be used either to train a classification network from scratch, or fine-tune it. We show theoretically why GQ encourages the observation made by ODIN and evaluate GQ on a number of network architectures and datasets. Experiments show that we achieve SOTA on a large number of standard benchmarks.



author    = {Amir Jevnisek and Shai Avidan},
title     = {Learning ODIN},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year      = {2022},
url       = {https://bmvc2022.mpi-inf.mpg.de/0210.pdf}

Copyright © 2022 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection