HACK_GAN_img2img

   

Based on pix2pix by Isola et al.

PAPER

Image-to-Image Translation with Conditional Adversarial Nets [CVPR 2017]

HACKX

Project URL

TEAM NAME

取个名字真TM难

AUTHOR

• XueWenLiao
• ChengChen
• LuoJie
• LiuDong
• WuQin Surprise Scene

PREREQUISITES

• Android 7.0
• Tensorflow 1.4.1
• Django 2.0.5
• djangorestframework 3.8.2

CODE

Client	Server

RECOMMENDED

• Linux with Tensorflow GPU edition + cuDNN

DESCRIPTION

Based on GAN technology, quickly transform your own ideas into images and improve communication efficiency. First, use Tensorflow to build GAN models. Second, encapsulate models into functions. Third, use Django framework to build servers and provide APIs for clients to use. The fourth use of Android to build a client. GAN neural network will be hand-drawn sketches of the user into a very realistic picture.

LOGO

PHOTOS

FEATURES DESCRIPTION

Painting:

1. Hand-Painted: users can freely go to paint, without any restrictions.
2. Eraser: erase the wrong part.
3. Line: users can only use the line to draw.
4. Revocation: Users can undo the previous step.
5. Empty: the user can empty the drawing board.
6. Generate a picture: Save the picture.

MODEL:

1. Building model: Users can use five label components: walls, doors, windows, eaves and room pillars to help draw.

2. Street View Model: Users can use five label components: roads, lawns, cars, trees, and street lights to help draw.

3. Package Model: the user draws the package.

4. Shoe Model: the user draws shoes.

VIEDO

DEMO URL

SLIDE DECK

Setup

Getting Started

step1:

# clone this repo
git clone [email protected]:luojie1024/HACK_GAN_img2img.git
cd HACK_GAN_MB

step2:

python manage.py runserver

Server Settings

DataSet

Please use the download_dataset.sh shell scripts to solve dataset can't be download problem!

dataset	example
python tools/download-dataset.py facades 400 images from CMP Facades dataset. (31MB) Pre-trained: BtoA
python tools/download-dataset.py cityscapes 2975 images from the Cityscapes training set. (113M) Pre-trained: AtoB BtoA
python tools/download-dataset.py maps 1096 training images scraped from Google Maps (246M) Pre-trained: AtoB BtoA
python tools/download-dataset.py edges2shoes 50k training images from UT Zappos50K dataset. Edges are computed by HED edge detector + post-processing. (2.2GB) Pre-trained: AtoB
python tools/download-dataset.py edges2handbags 137K Amazon Handbag images from iGAN project. Edges are computed by HED edge detector + post-processing. (8.6GB) Pre-trained: AtoB

The facades dataset is the smallest and easiest to get started with.