No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
data/play_store_json
tf-faster-rcnn
tools
LICENSE
README.md

README.md

Wirifying App Screenshots

An approach to automatically collect different types of UI components from screenshots of existing mobile apps.

This work leverages a Tensorflow implementation of Faster RCNN, mainly based on the work of endernewton.

Prerequisites

pip install Cython opencv-python easydict numpy scipy scikit-image six lxml Pillow imgaug tqdm

Getting Started

  1. Clone the repository
git clone https://github.com/bernardcwj/FYP_2017.git
  1. Update your -arch in setup script to match your GPU
cd FYP_2017/tf-faster-rcnn/lib
# Change the GPU architecture (-arch) if necessary
vim setup.py
GPU model Architecture
TitanX (Maxwell/Pascal) sm_52
GTX 960M sm_50
GTX 1080 (Ti) sm_61
Grid K520 (AWS g2.2xlarge) sm_30
Tesla K80 (AWS p2.xlarge) sm_37

Note: You are welcome to contribute the settings on your end if you have made the code work properly on other GPUs. Also even if you are only using CPU tensorflow, GPU based code (for NMS) will be used by default, so please set USE_GPU_NMS False to get the correct output.

  1. Build the Cython modules
make clean
make
cd ../..

Test with pre-trained model

  1. Download training and test data
cd data
wget ...
unzip androidAUG.zip
cd ..
  1. Create symlinks for the Android dataset
cd tf-faster-rcnn/data/VOCdevkit
ln -s  ../../../data/android_data_aug android_data
cd ../..
  1. Download and extract pre-trained model Android dataset with image augmentation here
unzip android_aug_res101_140k.zip
sudo cp -r android_aug_res101_140k/VOCdevkit data/
sudo cp -r android_aug_res101_140k/cache data/
sudo cp -r android_aug_res101_140k/output .
  1. Test with pre-trained model
./experiments/scripts/train_faster_rcnn.sh [GPU_ID] [DATASET] [NET]
# GPU_ID is the GPU you want to test on
# NET is the network arch to use
# DATASET is defined in train_faster_rcnn.sh
./experiments/scripts/test_faster_rcnn.sh 0 android_voc res101

Demo on app introductory screenshots crawled from Google Play App Store

A collection of app meta-data files can be found in data/play_store_json.

Note: For demonstration purposes, the data of only ten apps are made available

  1. Preprocessing Crawl images from Google Play App Store
# By default, crawled images are saved under data/play_store_screenshots
cd ..
python tools/google-play-screenshot-scraper.py 
  1. Run demo on crawled images
# By default, demo outputs are saved under demo_output
cd tf-faster-rcnn
./tools/demo_modified.py