Spot-the-difference self-supervised pre-training for anomaly detection and segmentation

Y Zou, J Jeong, L Pemula, D Zhang… - European Conference on …, 2022 - Springer
Visual anomaly detection is commonly used in industrial quality inspection. In this paper, we
present a new dataset as well as a new self-supervised learning method for ImageNet pre …

Contrastive and non-contrastive self-supervised learning recover global and local spectral embedding methods

R Balestriero, Y LeCun - Advances in Neural Information …, 2022 - proceedings.neurips.cc
Abstract Self-Supervised Learning (SSL) surmises that inputs and pairwise positive
relationships are enough to learn meaningful representations. Although SSL has recently …

Partial success in closing the gap between human and machine vision

R Geirhos, K Narayanappa, B Mitzkus… - Advances in …, 2021 - proceedings.neurips.cc
A few years ago, the first CNN surpassed human performance on ImageNet. However, it
soon became clear that machines lack robustness on more challenging test cases, a major …

Getting aligned on representational alignment

I Sucholutsky, L Muttenthaler, A Weller, A Peng… - ar** high-level visual representation in brains and machines?
C Conwell, JS Prince, KN Kay, GA Alvarez, T Konkle - BioRxiv, 2022 - biorxiv.org
The rapid development and open-source release of highly performant computer vision
models offers new potential for examining how different inductive biases impact …

Improving neural network representations using human similarity judgments

L Muttenthaler, L Linhardt, J Dippel… - Advances in …, 2023 - proceedings.neurips.cc
Deep neural networks have reached human-level performance on many computer vision
tasks. However, the objectives used to train these networks enforce only that similar images …