JobTopGun - Job Title Classification: Kaggle Competition

Achieving 5th place in Kaggle competition in developing a model to classify the job titles in both Thai and English into 597 occupation classes and submit them for evaluation, outperformed 64 other participants, including both graduates and undergraduates.

Speaker Diarization for Thai Speeches

Improved the speaker diarization model for the Thai language, achieving a 4x reduction in the diarization error rate compared to the pretrained model of pyannote. This improvement has resulted in a significant decrease from approximately 20% to around 5%.

Automated measurement of hip–knee–ankle angle in total knee arthroplasty using image-based raptor assistance

Develop executable Python software that takes lower limb radiograph images as input and provides automated measurements of the hip-knee-ankle angle. Additionally, the software labels the lines and angle values in the images, providing a comprehensive analysis of the lower limb alignment.

OCR Web application for Thai Handwriting

Develop a complete development and deployment of a web application. Our main focus was on creating an OCR system specifically designed for Thai Handwriting, utilizing deep learning techniques.

PM2.5 Time-series Prediction Model

Achieving 1st place in Kaggle competition in developing of a time-series analysis model, our goal was to forecast PM2.5 levels 72 hours in advance for six stations in each region. To enhance the accuracy of our model, we employed web scraping techniques to gather additional data relevant to PM2.5.