https://github.com/nwhovian/drcn-digits-classification

Domain Adaptation for digits classification using Deep Reconstruction-Classification Network
https://github.com/nwhovian/drcn-digits-classification

digits-classification domain-adaptation drcn mnist-dataset svhn-dataset

Last synced: 3 months ago
JSON representation

Domain Adaptation for digits classification using Deep Reconstruction-Classification Network

• Host: GitHub
• URL: https://github.com/nwhovian/drcn-digits-classification
• Owner: nWhovian
• License: mit
• Created: 2020-04-21T21:44:40.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2020-04-21T22:04:31.000Z (about 5 years ago)
• Last Synced: 2025-01-10T21:24:14.081Z (5 months ago)
• Topics: digits-classification, domain-adaptation, drcn, mnist-dataset, svhn-dataset
• Language: Jupyter Notebook
• Size: 8.79 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Domain Adaptation for Digits Classification

#### train on SVHN dataset, test on MNIST

For the domain adaptation I use the **Deep Reconstruction-Classification Network (DRCN)**.

![](https://imgur.com/5Ggxr8D.png)

The model is based on a convolutional architecture that has two pipelines with a shared encoding representation. First pipeline is a convolutional network for label prediction based on the source data, second pipeline is a convolutional autoencoder for target data reconstruction. Including the reconstruction of target data among with a standard label classifier helps to implement the domain adaptation.

The model is based on a [paper](https://arxiv.org/pdf/1607.03516.pdf) and a [code](https://github.com/fungtion/DRCN).

## Results

![](https://imgur.com/fsKTJcR.png)