Table of contents

• Table of contents
• Xlearn Transfer Learning Library
  • Setup
  • Usage
    • Quick Start
  • Methods

Xlearn Transfer Learning Library

The Xlearn transfer learning library implements some tranfer learning or domain adaption algorithms using TensorFlow. The script main.py is for both training and evaluating the models. And it also contains some scripts for downloading and preprocessing the data and the pretrained models.

This library is still being heavily developed. Vast changes are expected these days.

Setup

This library is Python 2&3 compatible. See requirements.txt for the detailed dependencies. You may modify it if you want to use the gpu version of TensorFlow.

Usage

Quick Start

Excute the following script can run Alexnet-based DAN models on the Office dataset.

# Prepare the data
sh data/office/download.sh
# Download and process the pretrained mean
python download_mean_from_caffe.py
# Download and process the pretrained model
python download_model_from_caffe.py
# Train
python main.py

Methods

Currently it only supports DAN with Caffe alexnet base-model.