Introduction

This is the development branch for Trans-Learn. Compared with the master version, we have added

• Regression DA （including Source Only, DD)
• Unsupervised DA (including MCC)
• Partial DA (DANN, PADA)
• Open Set DA (DANN, OSBP)
• Segmentation DA (ADVENT, FDA)

We are planning to add

• Segmentation DA (CycleGAN, Cycada)
• Keypoint Detection DA
• Finetune Library (ftlib)

Trans-Learn is a Transfer Learning library based on pure PyTorch with high performance and friendly API. Our code is pythonic, and the design is consistent with torchvision. You can easily develop new algorithms, or easily apply existing algorithms..

• Latest results for unsupervised DA can be found at Unsupervised DA benchmark
• Results for partial DA can be found at Partial DA benchmark
• Results for open-set DA can be found at Open-set DA benchmark
• Results for regression DA can be found at Regression DA benchmark
• A new documentation can be found at Documentation (please open in Firefox or Safari). Note that this link is only for temporary use.

There might be many errors and changes in this branch. Please refer master for stable version. Also, any suggestions are welcome!