Efficient Visual Pretraining with Contrastive Detection

TL;DR

This paper introduces a contrastive detection method for self-supervised visual pretraining that achieves high transfer accuracy with significantly less computation, making it more efficient than existing methods.

Contribution

The paper proposes a novel contrastive detection objective that reduces pretraining computational costs while maintaining state-of-the-art transfer performance.

Findings

Achieves state-of-the-art transfer accuracy with up to 10x less pretraining.

Performs on par with large-scale systems using 1000x less data.

Effectively handles complex images like COCO, closing the gap with supervised learning.

Abstract

Self-supervised pretraining has been shown to yield powerful representations for transfer learning. These performance gains come at a large computational cost however, with state-of-the-art methods requiring an order of magnitude more computation than supervised pretraining. We tackle this computational bottleneck by introducing a new self-supervised objective, contrastive detection, which tasks representations with identifying object-level features across augmentations. This objective extracts a rich learning signal per image, leading to state-of-the-art transfer accuracy on a variety of downstream tasks, while requiring up to 10x less pretraining. In particular, our strongest ImageNet-pretrained model performs on par with SEER, one of the largest self-supervised systems to date, which uses 1000x more pretraining data. Finally, our objective seamlessly handles pretraining on more complex images such as those in COCO, closing the gap with supervised transfer learning from COCO to PASCAL.