Inverse Problems Leveraging Pre-trained Contrastive Representations

Sriram Ravula, Georgios Smyrnis, Matt Jordan, Alex Dimakis

Keywords: Inverse Problems, Representation Learning, Contrastive Learning, Robustness

TL;DR: We obtain representations for highly corrupted images by using a supervised inversion method, which leverages contrastive learning.

Abstract: We study a new family of inverse problems for recovering representations of corrupted data. We assume access to a pre-trained representation learning network R(x) that operates on clean images, like CLIP. The problem is to recover the representation of an image R(x), if we are only given a corrupted version A(x), for some known forward operator A. We propose a supervised inversion method that uses a contrastive objective to obtain excellent representations for highly corrupted images. Using a linear probe on our robust representations, we achieve a higher accuracy than end-to-end supervised baselines when classifying images with various types of distortions, including blurring, additive noise, and random pixel masking. We evaluate on a subset of ImageNet and observe that our method is robust to varying levels of distortion. Our method outperforms end-to-end baselines even with a fraction of the labeled data in a wide range of forward operators.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

Code: https://github.com/Sriram-Ravula/Contrastive-Inversion

Community Implementations: [ 2 code implementations](https://www.catalyzex.com/paper/arxiv:2110.07439/code)