EasyTL: Practically Easy Transfer Learning

This directory contains the code for paper Easy Transfer Learning By Exploiting Intra-domain Structures published at IEEE International Conference on Multimedia & Expo (ICME) 2019.

Requirements

There are two implementations of EasyTL: Matlab and Python.

Matlab

The original code is written using Matlab R2017a. I think all versions after 2015 can run the code.

Python

Thanks to @KodeWorker for providing a Python implementation.

The Python version can be found in here.

Demo & Usage

For Matlab, I offer three basic demos to reproduce the experiments in this paper:

Note that this directory does not contains any dataset. You can download them at the following links, and then add the folder to your Matlab path before running the code.

For Python, the demo code is .py.

Download Amazon Review dataset.

Download Office-Caltech with SURF features

Download Image-CLEF ResNet-50 pretrained features

Download Office-Home ResNet-50 pretrained features

You are welcome to run EasyTL on other public datasets such as here. You can also use your own datasets.

Reference

If you find this code helpful, please cite it as:

Jindong Wang, Yiqiang Chen, Han Yu, Meiyu Huang, Qiang Yang. Easy Transfer Learning By Exploiting Intra-domain Structures. IEEE International Conference on Multimedia & Expo (ICME) 2019.

Or in bibtex style:

@inproceedings{wang2019easytl,
    title={Easy Transfer Learning By Exploiting Intra-domain Structures},
    author={Wang, Jindong and Chen, Yiqiang and Yu, Han and Huang, Meiyu and Yang, Qiang},
    booktitle={IEEE International Conference on Multimedia & Expo (ICME)},
    year={2019}
}

Results

EasyTL achieved state-of-the-art performances compared to a lot of traditional and deep methods as of March 2019: