Article(メインAtmaCup)

[ju-ki]

• 自己教師あり学習 https://www.guruguru.science/competitions/17/discussions/a39d588e-aff2-4728-8323-b07f15563552/

• TTA https://www.guruguru.science/competitions/17/discussions/4864ee81-6336-4cad-bb74-a4c9b46e6eb2/

• albumentationsの実装 https://github.com/Kazuhito00/albumentations-examples

• mask系のaugumentation https://ohke.hateblo.jp/entry/2020/06/27/230000

• jigsaw puzzleの自己教師あり学習? https://www.guruguru.science/competitions/17/discussions/fe12e2a4-d6ba-47c7-b2eb-740c409fd2f9/

• contrastive learning(自己教師あり学習) https://www.guruguru.science/competitions/17/discussions/1775e171-2e66-4823-a3de-943381339fd4/

[ju-ki]

• facebookによるsimsiamの実装 https://github.com/facebookresearch/simsiam

[ju-ki]

• 俵さんの解法が結構参考になる https://www.guruguru.science/competitions/17/discussions/a768baf7-a805-46bb-ae29-88c92fd30fe6/

[ju-ki]

• cutmixの実装(pytorch) https://www.kaggle.com/ar2017/pytorch-efficientnet-train-aug-cutmix-fmix

[ju-ki]

• augmixの実装(まだAlbumentationが出ていないらしいので各手法のスクラッチ版が見れるので結構いいかも) https://www.kaggle.com/kaushal2896/efficientnet-b1-with-gridmask-augmix

[ju-ki]

-grid maskの実装

https://www.kaggle.com/haqishen/gridmask