DenseCL

Dense Contrastive Learning for Self-Supervised Visual Pre-Training

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 (DenseCL), which implements self-supervised learning by optimizing a pairwise contrastive (dis)similarity loss at the pixel level between two views of input images.

Results and Models

This page is based on documents in MMSelfSup.

Classification

The classification benchmarks includes 4 downstream task datasets, VOC, ImageNet, iNaturalist2018 and Places205. If not specified, the results are Top-1 (%).

VOC SVM / Low-shot SVM

The Best Layer indicates that the best results are obtained from which layers feature map. For example, if the Best Layer is feature3, its best result is obtained from the second stage of ResNet (1 for stem layer, 2-5 for 4 stage layers).

Besides, k=1 to 96 indicates the hyper-parameter of Low-shot SVM.

Self-Supervised Config	Best Layer	SVM	k=1	k=2	k=4	k=8	k=16	k=32	k=64	k=96
r50_4xb64_cos_fp16_ep200	feature5	82.5	42.68	50.64	61.74	68.17	72.99	76.07	79.19	80.55

ImageNet Linear Evaluation

The Feature1 - Feature5 don't have the GlobalAveragePooling, the feature map is pooled to the specific dimensions and then follows a Linear layer to do the classification. Please refer to r50_mhead_sz224_4xb64_step_ep90.py for details of config.

The AvgPool result is obtained from Linear Evaluation with GlobalAveragePooling. Please refer to r50_linear_sz224_4xb64_step_ep100.py for details of config.

Self-Supervised Config	Feature1	Feature2	Feature3	Feature4	Feature5	AvgPool
r50_4xb64_cos_fp16_ep200	15.86	35.47	49.46	64.06	62.95	63.34

Places205 Linear Evaluation

The Feature1 - Feature5 don't have the GlobalAveragePooling, the feature map is pooled to the specific dimensions and then follows a Linear layer to do the classification. Please refer to r50_mhead_sz224_4xb64_step_ep28 for details of config.

Self-Supervised Config	Feature1	Feature2	Feature3	Feature4	Feature5
r50_4xb64_cos_fp16_ep200	21.32	36.20	43.97	51.04	50.45

Detection

The detection benchmarks includes 2 downstream task datasets, Pascal VOC 2007 + 2012 and COCO2017. This benchmark follows the evluation protocols set up by MoCo.

Pascal VOC 2007 + 2012

Please refer to faster_rcnn_r50_c4_mstrain_24k_voc0712.py for details of config.

Self-Supervised Config	AP50
r50_4xb64_cos_fp16_ep200	82.14

COCO2017

Please refer to mask_rcnn_r50_fpn_mstrain_1x_coco.py for details of config.

Segmentation

The segmentation benchmarks includes 2 downstream task datasets, Cityscapes and Pascal VOC 2012 + Aug. It follows the evluation protocols set up by MMSegmentation.

Pascal VOC 2012 + Aug

Please refer to fcn_r50-d8_512x512_20k_voc12aug.py for details of config.

Self-Supervised Config	mIOU
r50_4xb64_cos_fp16_ep200	69.47

Citation

@inproceedings{wang2021dense,
  title={Dense contrastive learning for self-supervised visual pre-training},
  author={Wang, Xinlong and Zhang, Rufeng and Shen, Chunhua and Kong, Tao and Li, Lei},
  booktitle={CVPR},
  year={2021}
}