https://github.com/nashory/pixel-level-dt-torch

Torch implementation of "Pixel-Level Domain Transfer"
https://github.com/nashory/pixel-level-dt-torch

domain-transfer generative-adversarial-network

Last synced: 12 days ago
JSON representation

Torch implementation of "Pixel-Level Domain Transfer"

• Host: GitHub
• URL: https://github.com/nashory/pixel-level-dt-torch
• Owner: nashory
• Created: 2017-10-27T03:11:03.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-11-06T01:52:39.000Z (over 7 years ago)
• Last Synced: 2024-12-02T00:04:20.778Z (2 months ago)
• Topics: domain-transfer, generative-adversarial-network
• Language: Lua
• Homepage:
• Size: 45.9 KB
• Stars: 4
• Watchers: 3
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# pixel-level-dt-torch

Torch implementation of "[Pixel-Level Domain Transfer](https://arxiv.org/pdf/1603.07442)", bug-fixed version of [repo](https://github.com/fxia22/PixelDTGAN)

![image](https://github.com/fxia22/pldtgan/raw/master/gan.jpg)

## Note (IMPORTANT):

+ I found too many bugs in the [original repo](https://github.com/fxia22/PixelDTGAN), and the code was not running.

+ Finally, I decided to release __bug-fixed__ version of "[Pixel-Level Domain Transfer](https://arxiv.org/pdf/1603.07442)" with enhanced features.

+ The majority of the codes were brought from [original repo](https://github.com/fxia22/PixelDTGAN).

## What is different from original repo?

+ Most importantly, BUG FIXED.

+ The input sizes of both 64 x 64 to 128 x 128 are supported.

+ Network structure was slightly modified acccording to input size.

+ A grid form of generated images are saved in png format.

+ Torch logger is added.

## Prerequisites

+ [Torch7](http://torch.ch/docs/getting-started.html#_)

+ [display](https://github.com/szym/display)

## Dataset preparation

+ (step 1) Download LOOKBOOK dataset: [Here](https://drive.google.com/file/d/0By_p0y157GxQU1dCRUU4SFNqaTQ/view?usp=sharing)

+ (step 2) Place 'lookbook.tar' zip file at root dir.

+ (step 3) run `sh setup.sh`

+ (step 4) preprocess dataset using `prepare_data.ipynb` (you may need to use jupyter notebook)

## How to run?

+ Start training:

~~~

(change training options in opts.lua file beforehand)

python run.py

python run.py & (if want to run in background)

~~~

+ Run server for visualization:

~~~

(change server_ip and server_port options in opts.lua file beforehand)

th server.lua

th server.lua & (if want to run in background)

~~~

## Visualization (display)  

You can see the generated images and loss graphs using web browser.  

`https://:`

(example)

![image](https://puu.sh/y8eZp/53c12325b2.png)

## Experimental Results

Result so far (it is still being trained at this moment.)

|training|Final|  

|---|---|  

| condition: 128x128, 0.8 epoch | condition: 128x128, 0.8 epoch |

|||

## [!!] Trouble-shooting Multi-GPU Memory Allocation issue in torch

It seems that torch uses all memories across the GPUs if you use muti gpu(https://github.com/torch/cutorch/issues/180).

To prevent this and use only single GPU, I gave 'UDA_VISIBLE_DEVICES=n' options when running lua script.

If you are running with single GPU, or do not want to disable multi-gpu memory allocation of torch, remove this option in `run.py`

__(e.g.) os.system('CUDA_VISIBLE_DEVICES=0 th ./main.lua') --> os.system('th ./main.lua')__

## Acknowledgement

+ [@fxia22's original repo](https://github.com/fxia22/PixelDTGAN)

## Author

MinchulShin / [@nashory](https://github.com/nashory)  

__Any insane bug reports or questions are welcome. (min.stellastra[at]gmail.com)  :-)__