Revisiting Weakly Supervised Pre-Training of Visual Perception Models

TL;DR

This paper demonstrates that weakly supervised pre-training using hashtags can outperform self-supervised methods in visual recognition tasks, offering a promising alternative to traditional fully supervised approaches.

Contribution

It introduces SWAG, a weakly supervised pre-training method using hashtags, and shows its competitive performance against self-supervised models across various transfer-learning settings.

Findings

Weakly supervised models outperform self-supervised counterparts.

SWAG achieves strong transfer-learning performance.

Models do not learn harmful stereotypes.

Abstract

Model pre-training is a cornerstone of modern visual recognition systems. Although fully supervised pre-training on datasets like ImageNet is still the de-facto standard, recent studies suggest that large-scale weakly supervised pre-training can outperform fully supervised approaches. This paper revisits weakly-supervised pre-training of models using hashtag supervision with modern versions of residual networks and the largest-ever dataset of images and corresponding hashtags. We study the performance of the resulting models in various transfer-learning settings including zero-shot transfer. We also compare our models with those obtained via large-scale self-supervised learning. We find our weakly-supervised models to be very competitive across all settings, and find they substantially outperform their self-supervised counterparts. We also include an investigation into whether our models learned potentially troubling associations or stereotypes. Overall, our results provide a compelling argument for the use of weakly supervised learning in the development of visual recognition systems. Our models, Supervised Weakly through hashtAGs (SWAG), are available publicly.