/Abacus.AI

Applying computer vision for simple math problems

Primary LanguagePython

AbacusLogo

Abacus.AI

Description: Detect and classify basic mathematical elements on an image and evaluate expressions based on the same

Project for bachelors thesis at University of Zagreb, Faculty of Electrical Engineering and Computing

Full thesis name: Detection and classification of text based elements on an image using deep neural networks
Module: Computer Science
Mentor: prof. dr. sc. Domagoj Jakobović

Contents:

  • Pipeline for creating training images and corresponding .csv files
  • Handwritten fonts
  • All symbols generated in 100+ fonts
    • Sorted by difficulty
  • Image examples
  • Heavily documented code
    • Code will be made more efficient after presenting the final thesis
  • generate_tfrecords.py for creating *.record files for further Tensorflow API usage

Testing:

  • File called ImagePredictor.py contains everything neccessary to run the interference graph on an image
    • Output image is saved as output.png and the result is printed out in the console