Dense Contrastive Learning for Self-Supervised Visual Pre-Training

TL;DR

This paper introduces dense contrastive learning, a self-supervised approach that optimizes pixel-level feature correspondence, significantly improving performance on dense prediction tasks with minimal additional computational cost.

Contribution

It proposes a novel dense contrastive learning method that directly aligns local features at the pixel level, outperforming existing methods like MoCo-v2 on various dense prediction benchmarks.

Findings

Achieves up to 2.0% AP improvement on PASCAL VOC object detection

Outperforms state-of-the-art methods on multiple dense prediction tasks

Adds less than 1% computational overhead compared to MoCo-v2

Abstract

To date, most existing self-supervised learning methods are designed and optimized for image classification. These pre-trained models can be sub-optimal for dense prediction tasks due to the discrepancy between image-level prediction and pixel-level prediction. To fill this gap, we aim to design an effective, dense self-supervised learning method that directly works at the level of pixels (or local features) by taking into account the correspondence between local features. We present dense contrastive learning, which implements self-supervised learning by optimizing a pairwise contrastive (dis)similarity loss at the pixel level between two views of input images. Compared to the baseline method MoCo-v2, our method introduces negligible computation overhead (only <1% slower), but demonstrates consistently superior performance when transferring to downstream dense prediction tasks including object detection, semantic segmentation and instance segmentation; and outperforms the state-of-the-art methods by a large margin. Specifically, over the strong MoCo-v2 baseline, our method achieves significant improvements of 2.0% AP on PASCAL VOC object detection, 1.1% AP on COCO object detection, 0.9% AP on COCO instance segmentation, 3.0% mIoU on PASCAL VOC semantic segmentation and 1.8% mIoU on Cityscapes semantic segmentation. Code is available at: https://git.io/AdelaiDet