/SimSiam

Exploring Simple Siamese Representation Learning

Primary LanguagePythonMIT LicenseMIT

Exploring Simple Siamese Representation Learning (SimSiam)

Network

simsiam

Experiments

Model Pre-training Epochs Batch size Dim Linear Evaluation Acc (%)
ResNet-18 (Paper) 800 512 2048 O 91.8
ResNet-18 (Our) 300 512 1024 O 72.49
ResNet-18 800 256 1024 O 83.93
ResNet-18 512 2048 O wip
  • plot

Usage

  • Dataset (CIFAR-10)
    data
      └── cifar-10-batches-py
        ├── batches.meta
        ├── data_batch_1
        ├── data_batch_2
        ├── data_batch_3
        ├── data_batch_4
        ├── data_batch_5
        ├── readme.html
        └── test_batch
    
  1. Pre-training
python main.py --pretrain True
  1. DownStream Task (Linear)
python main.py --checkpoints checkpoints/checkpoint_pretrain_model.pth --pretrain False

Reference

  • Paper Link
  • Author: Xinlei Chen, Kaiming He
  • Organization: Facebook AI Research (FAIR)