Font Classifier

This is the font-classifier project for the computer vision course. To read more about the project, read the report located here.

How to run the model and generate the test results?

1. Install all the dependencies:
   • numpy
   • matplotlib
   • h5py
   • pytorch (Pay attention to install the CUDA version. The code will also work without it, but won't finish in reasonable time)
   • open-cv
2. Download the pre-trained model from this link, unzip it, and replace the empty file in models/all_models_without_perms with it.
3. Download the test data from here and replace the empty file in Project - Test Set\SynthText_test.h5 with it.
4. Run the model_testing.py file.

File structure

• The Project and Project - Test Set folders contains a template for the train and test datasets. There are empty files that should be replaced with the real train and test datasets.
• The models folder is a templaye for the pre-trained model. There's an empty file that should be replaced with the real trained model weights. Any other model that will be trained will be located in this folder.
• The outputs folder contains the output of the model_training.py script when the pre-trained model was trained. Any other model that will be trained will be located in this folder.
• The report folder is the most important one - it contains the report and the test labels in the required csv file format.