Skip to content

Latest commit

 

History

History
28 lines (17 loc) · 1.87 KB

README.md

File metadata and controls

28 lines (17 loc) · 1.87 KB
Raw

FAR - Finite Automaton Recognizer

This is a repository for my bachelor thesis, the project aims to allow the recognition of handrawn finite automata.

How to Use

  1. Download the models: https://drive.google.com/drive/folders/17uZaehJ-_TW_3oGN3rMZi77hRcoy6DNG?usp=sharing and extract the contents of the archive into a directory /FAR/recognition/models so that the model files are directly under this directory (/FAR/recognition/models/{model}.h5)

  2. Have a working python environment installed and also install the python docker module pip install docker

  3. Navigate to the repository root: (/FAR)

  4. Build the container: docker build . -t far (this takes a long time)

  5. Navigate to the recognition directory: (/FAR/recognition)

(optional) Do a test run with one of the test images: python3 recognize.py ../test_images/automaton1.png

  1. Add your images to the IO directory (create this directory if needed): (/FAR/recognition/IO) (sometimes depending on the OS it is necessary to add permissions on this folder, e.g. chmod 777 IO)

  2. Execute the recognition script: python3 recognize.py IO/{your_image}.png (if you have spaces in your image file name, escape them and enclose the path in ticks e.g. IO/contains spaces image.png -> 'IO/contains\ spaces\ image.png'

After the script is executed, your default browser should open and show the recognition result. If your browser does not open, navigate to the IO directory and open the 0_frontend.html yourself. Default browsers on different OS make this makeshift frontend implementation difficult. Firefox may also cause problems so use Chromium based browsers as a last resort.

The IO directory gets mounted into the docker container which is the reason why we add input images here. After running the tool it contains all the intermediary results which are mentioned in the thesis.