LIVIAETS/KD_UDA

KD-UDA

This is the repository for Joint Progressive Knowledge Distillation andUnsupervised Domain Adaptation.

Requirements

• pytorch=>1.1.0
• visdom
• https://github.com/luizgh/visdom_logger

Datasets:

• Office31: https://people.eecs.berkeley.edu/~jhoffman/domainadapt/#datasets_code
• ImageClef-DA: https://www.imageclef.org/2014/adaptation

How to run:

One you have installed all the requirements and download the dataset. Please take a look at the file kd_da_alt.py. All the hyper-parameters can be found in the main function(). By default we assume the dataset to be at, for example webcam: "~/datasets/webcam/images"

Once you have configured all paths, you can use: python kd_da_alt.py to launch a training

Acknowledgements

We used some of the code from:

• https://github.com/thuml/CDAN: to load ImageClef dataset
• https://github.com/jindongwang/transferlearning/tree/master/code/deep/TCP: to load Office31 dataset and MMD loss