https://github.com/ChillingDream/Lifelong-Gan

A tensorlayer implementation of LifeLong Gan
https://github.com/ChillingDream/Lifelong-Gan

gan

Last synced: 2 months ago
JSON representation

A tensorlayer implementation of LifeLong Gan

• Host: GitHub
• URL: https://github.com/ChillingDream/Lifelong-Gan
• Owner: ChillingDream
• License: mit
• Created: 2019-10-24T07:26:28.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2020-01-13T11:50:45.000Z (about 5 years ago)
• Last Synced: 2024-08-02T20:44:31.567Z (6 months ago)
• Topics: gan
• Language: Python
• Homepage:
• Size: 231 KB
• Stars: 13
• Watchers: 2
• Forks: 5
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-tensorlayer - Lifelong GAN

README

        
# LifeLong-Gan

### Requirements

* Tensorflow 2.0.0

* Tensorlayer 2.0

* Python 3.6

* Numpy

* Tqdm

### Model



### Prepare data

* Create a folder `data` in the project directory.

* You can run `download_dataset.sh` in folder datasets to download the images or directly put the images in folder `data`, training data in a subdirectory `train` and test data in `val`.

### Run

* Training

  ``` bash

  python train.py --tasks edges2shoes+facades

  ```

  Tasks edges2shoes and facades will be trained in turn. You can replace them with your datasets and concatenate the task names with "+". More training settings can be found in `params.py`.

* Evaluation

  ```bash

  python evaluate.py --tasks edges2shoes+facades

  ```

  The evaluation results can be found in subdirectory `samples`.

  

### Reference

* [1] Lifelong GAN: Continual Learning for Conditional Image Generation. ICCV, 2019

* [2] Toward Multimodal Image-to-Image Translation. NeurIPS, 2017.