Improving Transferability of Representations via Augmentation-Aware Self-Supervision
Accepted to NeurIPS 2021
TL;DR: Learning augmentation-aware information by predicting the difference between two augmented samples improves the transferability of representations.
Dependencies
conda create -n AugSelf python=3.8 pytorch=1.7.1 torchvision=0.8.2 cudatoolkit=10.1 ignite -c pytorch
conda activate AugSelf
pip install scipy tensorboard kornia==0.4.1 sklearn
Checkpoints
We provide ImageNet100-pretrained models in this Dropbox link.
Pretraining
We here provide SimSiam+AugSelf pretraining scripts. For training the baseline (i.e., no AugSelf), remove --ss-crop and --ss-color options. For using other frameworks like SimCLR, use the --framework option.
STL-10
CUDA_VISIBLE_DEVICES=0 python pretrain.py \
--logdir ./logs/stl10/simsiam/aug_self \
--framework simsiam \
--dataset stl10 \
--datadir DATADIR \
--model resnet18 \
--batch-size 256 \
--max-epochs 200 \
--ss-color 1.0 --ss-crop 1.0
ImageNet100
python pretrain.py \
--logdir ./logs/imagenet100/simsiam/aug_self \
--framework simsiam \
--dataset imagenet100 \
--datadir DATADIR \
--batch-size 256 \
--max-epochs 500 \
--model resnet50 \
--base-lr 0.05 --wd 1e-4 \
--ckpt-freq 50 --eval-freq 50 \
--ss-crop 0.5 --ss-color 0.5 \
--num-workers 16 --distributed
Evaluation
Our main evaluation setups are linear evaluation on fine-grained classification datasets (Table 1) and few-shot benchmarks (Table 2).
linear evaluation
CUDA_VISIBLE_DEVICES=0 python transfer_linear_eval.py \
--pretrain-data imagenet100 \
--ckpt CKPT \
--model resnet50 \
--dataset cifar10 \
--datadir DATADIR \
--metric top1
few-shot
CUDA_VISIBLE_DEVICES=0 python transfer_few_shot.py \
--pretrain-data imagenet100 \
--ckpt CKPT \
--model resnet50 \
--dataset cub200 \
--datadir DATADIR
196 Dec 25, 2022
360 Dec 10, 2022
66 Dec 04, 2022
2k Jan 08, 2023
1 Oct 24, 2021
83 Dec 21, 2022
64 Dec 16, 2022
205 Dec 29, 2022
83 Dec 30, 2022
890 Jan 04, 2023
53 Nov 13, 2022
20 Jun 21, 2022
29 Oct 09, 2022
224 Dec 08, 2022
6 Nov 22, 2022
29 Nov 08, 2022
21 Sep 08, 2022
238 Nov 25, 2022
127 Dec 12, 2022
129 Dec 22, 2022