Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision

TL;DR

This paper demonstrates that large-scale self-supervised training on uncurated images enhances model robustness, fairness, and bias reduction, capturing diverse semantic and stylistic information without supervision.

Contribution

It introduces a method of training massive models on uncurated images without supervision, leading to more robust and fair models that learn diverse, salient visual information.

Findings

Models trained on uncurated images outperform supervised models in fairness and robustness.

The approach captures artistic styles, geolocations, and multilingual embeddings from visual content.

The resulting models are less biased and more equitable across diverse benchmarks.

Abstract

Discriminative self-supervised learning allows training models on any random group of internet images, and possibly recover salient information that helps differentiate between the images. Applied to ImageNet, this leads to object centric features that perform on par with supervised features on most object-centric downstream tasks. In this work, we question if using this ability, we can learn any salient and more representative information present in diverse unbounded set of images from across the globe. To do so, we train models on billions of random images without any data pre-processing or prior assumptions about what we want the model to learn. We scale our model size to dense 10 billion parameters to avoid underfitting on a large data size. We extensively study and validate our model performance on over 50 benchmarks including fairness, robustness to distribution shift, geographical diversity, fine grained recognition, image copy detection and many image classification datasets. The resulting model, not only captures well semantic information, it also captures information about artistic style and learns salient information such as geolocations and multilingual word embeddings based on visual content only. More importantly, we discover that such model is more robust, more fair, less harmful and less biased than supervised models or models trained on object centric datasets such as ImageNet.