# drcn

**Repository Path**: vinceyzw/drcn

## Basic Information

- **Project Name**: drcn
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-05-28
- **Last Updated**: 2021-05-28

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation (DRCN)

This code is an implementation of the DRCN algorithm presented in [1].

[1] M. Ghifary, W. B. Kleijn, M. Zhang, D. Balduzzi, and W. Li. ["Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation (DRCN)"](https://arxiv.org/abs/1607.03516), European Conference on Computer Vision (ECCV), 2016

Contact:
```
Muhammad Ghifary (mghifary@gmail.com)
```

## Requirements
* Python 2.7
* Tensorflow-1.0.1
* Keras-2.0.0
* numpy
* h5py

## Usage
To run the experiment with the (grayscaled) SVHN dataset as the source domain and the MNIST dataset as the target domain
```
python main_sm.py
```

The core algorithm is implemented in __drcn.py__.
Data augmentation and denoising strategies are included as well.

## Results
The source to target reconstruction below (SVHN as the source) indicates the successful training of DRCN.

```
python reconstruct_images.py
```

![alt text](https://github.com/ghif/drcn/blob/master/rec_src.png "Source to Target Reconstruction")

The classification accuracies of one DRCN run are plotted as follows -- the results may vary due to the randomness:

```
python plot_results.py
```

![alt text](https://github.com/ghif/drcn/blob/master/svhn-mnist_plot.png "Accuracy Plot")