Skip to content

🏁 Flag recognition neural network + flag images dataset

License

Notifications You must be signed in to change notification settings

iamvukasin/flagnet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

50 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Flagnet

Build Status License MIT

Flagnet is a neural network for detecting country flags in photos. The detection is backed up by our very own dataset – the collection of photos of flags for 193 United Nations member countries.

Table of contents

Getting started

Flagnet is built in Python 3. Before cloning the project, make sure that you have downloaded and installed Python and Pipenv.

After downloading Python and Pipenv, you have to download all dependencies via Pipenv:

$ pipenv install

Now you're ready for the project!

Configuration

All project configuration is controlled from a single place – the config.py file. Currently you can change these parameters:

MIN_IMAGE_SIZE = 416, 416         # minimum size of images in the dataset
SCREENSHOT_IMAGE_SIZE = 800, 600  # size of generated images from Flagwaver website
NUM_DOWNLOAD_WORKERS = 8          # number of parallel workers when downloading the dataset

Dataset

The images which are part of the dataset are stored in the dataset folder and organized into the folders by country ISO 3166-1 alpha-2 codes. Inside every folder, there is a credits.yml (and its visual Markdown representation credits.md) which contains a list of images with its author name, license and download URL. To download the dataset, position yourself in the root of the project and run the downloader:

$ python -m download.download_dataset

Contributing

Pull requests are welcome for both the dataset and the neural network.

License

Flagnet is released under the MIT license.

Releases

No releases published

Sponsor this project

Packages

No packages published

Languages