manojll / brahmilator

Systematic solution to translate early Brahmi Inscription context through computer vision and machine learning techniques.

Hi 👋, I'm Manoj Rathnapriya!

Code is my language, and technology is my canvas.

• 🏫 BEng (Hons) Software Engineering - University Of Westminster

• 📫  How to reach me: Linkedin

Connect With Me

Languages and Tools

📚  Latest Medium Articles

• GSoC’24 Week 03: Updates from My GSoC Journey)
• GSoC’24 Week 02: My First Task In GSoC
• GSoC Week 1: Kickoff With An Amazing Orientation Session
• The art of code reviewing
• The first step to Machine learning
• REST API using Spring Boot and PostgreSQL with my learning experience

➡️  More Articles...

📊 GitHub Stats:

Visitors Count

brahmilator's Issues

Optimize pre process module

Execute and evaluate what's the most suitable morph shape for each operation.

Implement a NLP module

Initial version of the Brahmilator maps words respectively with the database and shows the output but it can’t perform accurate and meaningful outputs in everytime. Due to the damages and broken parts of the inscriptions there's a lot of issues occurred when generating the sentences.

Mongo atlas optimizing with new words

Expanded language select drop-down menu alignment need to be fixed

test app for sample plates

Complete documentation

Fix 'No exception type(s) specified' issue in brahmi_backend\controller.py

CodeFactor found an issue: No exception type(s) specified

It's currently on:
brahmi_backend\controller.py:133

An out of index error occurs in segmentation model

Restructure repo

Model accuracy dropped in last iteration

Add validator to check input image is Valid Inscription or not

Core model needed to be modified

Got overfitted, re-trained it but still has some issues.

Creating a Character Matrix Algorithm to predict damage and broken letters

This was a crucial function that authors want to implement during the initial version of the project. The module should perform effectively while predicting the damage letters.

Optimizing front-end

Optimize the Image Recognizing core module

Currently project has the MobileNetv2 to recognize the extracted Brahmi images. The model has over 80+ accuracy but still need to be fine-tuned to perform in various conditions.

Fix 'Possible binding to all interfaces.' issue in brahmi_backend\controller.py

CodeFactor found an issue: Possible binding to all interfaces.

It's currently on:
brahmi_backend\controller.py:139

Add README.md and add instructions