Inverse Problems Leveraging Pre-trained Contrastive Representations

TL;DR

This paper introduces a supervised contrastive inversion method leveraging pre-trained representations to recover clean image features from corrupted data, outperforming traditional end-to-end models across various distortions.

Contribution

The paper presents a novel contrastive inversion approach that enhances representation recovery from corrupted images using pre-trained models like CLIP, improving classification accuracy with less labeled data.

Findings

Outperforms end-to-end baselines on corrupted image classification.

Robust to various distortions like noise, blurring, and masking.

Effective with limited labeled data across different forward operators.

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.