SimSiam

Exploring Simple Siamese Representation Learning

Abstract

Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our “SimSiam” method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning.

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_lr0_05_ep100	feature5	84.64	39.65	49.86	62.48	69.50	74.48	78.31	81.06	82.56
r50_4xb64_cos_lr0_05_ep200	feature5	85.20	39.85	50.44	63.73	70.93	75.74	79.42	82.02	83.44

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_lr0_05_ep100	16.27	33.77	45.80	60.83	68.21	68.28
r50_4xb64_cos_lr0_05_ep200	15.57	37.21	47.28	62.21	69.85	69.84

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_lr0_05_ep100	21.32	35.66	43.05	50.79	53.27
r50_4xb64_cos_lr0_05_ep200	21.17	35.85	43.49	50.99	54.10

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_lr0_05_ep100	79.80
r50_4xb64_cos_lr0_05_ep200	79.85

COCO2017

Please refer to mask_rcnn_r50_fpn_mstrain_1x_coco.py for details of config.

Self-Supervised Config	mAP(Box)	AP50(Box)	AP75(Box)	mAP(Mask)	AP50(Mask)	AP75(Mask)
r50_4xb64_cos_lr0_05_ep100	38.6	57.6	42.3	34.6	54.8	36.9
r50_4xb64_cos_lr0_05_ep200	38.8	58.0	42.3	34.9	55.3	37.6

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_lr0_05_ep100	48.35
r50_4xb64_cos_lr0_05_ep200	46.27

Citation

@inproceedings{chen2021exploring,
  title={Exploring simple siamese representation learning},
  author={Chen, Xinlei and He, Kaiming},
  booktitle={CVPR},
  year={2021}
}