I am Min Si Thu, a builder of AI models in Myanmar. The time of writing this article is 2024, January 1.
I made a lot of contributions in the field of artificial intelligence in 2023. Some of the projects are great. I feel I should list the projects as an article so that everyone can use and continue the projects. So, here is the list.
MyanmarGPT
Project — https://huggingface.co/jojo-ai-mst/MyanmarGPT
This one is my favorite project. Nearly every NLP researcher in Myanmar was focusing on BERT architecture, I feel like we should go to GPT — Generative Pretrained Transformer architecture. The main purpose is that both end-user and machine-learning scientists can use the model. For end-users, it is just using. People in ML can use tokenizer and fine-tuning. Fine-tuning the MyanmarGPT model makes it easier to build a custom Myanmar language model than using alternative language models. And it is free to use on hugging face.
Burmese Voice Recognizer Project
Documentation — https://burmese-voice.vercel.app/
Github — https://github.com/MinSiThu/burmese-voice
Burmese Voice Recognizer is a javascript AI library to be used by developers. It can be easily installed via npm. Currently, it can recognize 6 Burmese voices and background noises. The library can be used for all purposes (educational, commercial, personal). I hope I can release version 2 in 2024.
Myanmar AI Tutor
Project — https://myanmar-ai-tutor.streamlit.app/
Facebook page — https://www.facebook.com/myanmaraitutor
Myanmar AI Tutor is an application for end-users to freely ask in the Burmese Language. It is powered by a small language model. There is a lot of journey to go on this project.
Hospital Antibiotics Usage Dataset
Dataset — https://www.kaggle.com/datasets/minsithu/hospital-antibiotics-usage
Drug resistance to antibiotics is increasing around the world every day. Generations of antibiotics are dynamically changing in treatment. When I was a medical student in 2019, the students had to take part in a research competition. I and my classmates had a project to analyze the antibiotics usage at a 300-bed teaching hospital in Mandalay. We won the first prize. After these years, Myanmar has been in civil war since 2021. That antibiotics usage dataset should be released for further studies by doctors, and researchers all around the world, including Myanmar.
Audio Noise Dataset
Dataset — https://www.kaggle.com/datasets/minsithu/audio-noise-dataset This dataset is a part of the Burmese voice javascript library. I open-sourced this dataset so that people who wanna go for the audio project don’t have to record the noise from the environment. Audio datasets include ten types of noise.
MMDeepSnake
Documentation — https://mmdeepsnake.vercel.app/
Model — https://huggingface.co/jojo-ai-mst/burmese_snake_classifier
Since the start of the coup, Myanmar has been in chaos and people in rural regions are facing deadly snake-bite problems. The project is aimed to classify Burmese snake species using the Convolutional Neural Network. This is my very first project in artificial intelligence.
Podoco — Plant Disease Classifier App
Project — https://github.com/MinSiThu/Podoco
This project is to use tflite model in Android by classifying the casava tree plant diseases. This model is trained to classify an input image into one of 6 cassava disease classes: Bacterial Blight, Brown Streak Disease, Green Mite, Mosaic Disease, Healthy, and Unknown. The training dataset is curated by the Mak-AI team at Makerere University.
Since the 2021 coup, people in Myanmar are facing difficulties. No internet connections, electricity off for more than 16 hours a day, and armed conflicts every day. But artificial intelligence and innovations have to be continued by young people.
I hope I can make another list for 2024 too. 2024 could be a tough year. We are also becoming tough people.
Min Si Thu is someone who grew up playing with mud and sand on the roads of a small town.
