Shortcuts

SimSiam

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.

How to use it?

from mmpretrain import inference_model

predict = inference_model('resnet50_simsiam-100e-pre_8xb512-linear-coslr-90e_in1k', 'demo/bird.JPEG')
print(predict['pred_class'])
print(predict['pred_score'])

Models and results

Pretrained models

Model

Params (M)

Flops (G)

Config

Download

simsiam_resnet50_8xb32-coslr-100e_in1k

38.20

4.11

config

model | log

simsiam_resnet50_8xb32-coslr-200e_in1k

38.20

4.11

config

model | log

Image Classification on ImageNet-1k

Model

Pretrain

Params (M)

Flops (G)

Top-1 (%)

Config

Download

resnet50_simsiam-100e-pre_8xb512-linear-coslr-90e_in1k

SIMSIAM 100-Epochs

25.56

4.11

68.30

config

model | log

resnet50_simsiam-200e-pre_8xb512-linear-coslr-90e_in1k

SIMSIAM 200-Epochs

25.56

4.11

69.80

config

model | log

Citation

@inproceedings{chen2021exploring,
  title={Exploring simple siamese representation learning},
  author={Chen, Xinlei and He, Kaiming},
  booktitle={CVPR},
  year={2021}
}
Read the Docs v: latest
Versions
latest
stable
mmcls-1.x
mmcls-0.x
dev
Downloads
epub
On Read the Docs
Project Home
Builds

Free document hosting provided by Read the Docs.