Egyptian Hieroglyph Classification using ConvNet

This is a PyTorch implementation of the paper "A Deep Learning Approach to Ancient Egyptian Hieroglyphs Classification" by Barucci et al.

Please download the dataset "EgyptianHieroglyphDataset_Original" at my Google Drive.
Please download the dataset "EgyptianHieroglyphDataset_Original_Clean" (40 classes) at my Google Drive.

Image
Gardener Label	D21	E34	V31

Steps for running full Python code:

1. Download "EgyptianHieroglyphDataset_Original" dataset from my Google Drive
2. Download "src" folder in this repo
3. Install all the requirements for Python packages
4. Run main.py
5. The training will start right away!

Steps for running Jupyter Notebook:

1. Click Egyptian_model_with_ResNet_Modular.ipynb in my repo
2. Click "Open in Colab"
3. Download "EgyptianHieroglyphDataset_Original" dataset from my Google Drive and store it into your Google Drive
4. Connect the Google Colab with your Google Drive and run the codes
5. The training will start right away!

Performance (Accuracy on 40 classes):

1. ResNet-50: 98.6%

Prior implementations:

1. GlyphReader by Morris Franken which extracts features using Inception-v3 and classifies hieroglyphs using SVM.

TODO:

1. Implementation for Glyphnet
2. Implementation for Inception-v3
3. Implementation for Xception