SHARTAZ KHAN AKASH
A PERSONAL PORTFOLIO & ACADEMIC RECORD
Research Interests
Generative AI & LLMs
Exploring architecture and fine-tuning of Large Language Models and diffusion models.
Explainable AI (XAI)
Developing interpretable models to understand and trust complex AI systems.
Multimodal Learning
Fusing diverse data modalities like text, images, and sensor data for context-aware AI.
AI for Scientific Discovery
Leveraging ML to accelerate research in fields like materials science and drug discovery.
Publications
Action Recognition Based Real-time Bangla Sign Language Detection
Developed a real-time system to generate Bangla sentences from sign language video feeds using Blaze Pose and an LSTM model, achieving 87.14% validation accuracy.
ICREST 2023. View Publication →
Deep Learning-Based Prediction of Football Players' Performance
Created a system using YOLOv4, OpenCV, and pose estimation to predict goalpost targets during penalty shootouts by training an LSTM model on over 200,000 pose landmarks.
ICCIT 2023. View Publication →
Projects
Deep Learning for Football Penalty Prediction
Problem: Developed a system to predict the goalpost region targeted by players during penalty shootouts, addressing a complex challenge in sports analytics.
Methodology: Utilized YOLOv4 for player detection, OpenCV for image processing, and pose estimation to extract key body landmarks. Trained an LSTM model on over 200,000 pose landmarks to predict kicker actions.
Testimonials
"He is a multi-talented student poised to contribute to any research problem he takes on significantly. I wholeheartedly recommend him."
"He represents the caliber of students who do not just absorb knowledge but actively transform intellectual landscapes. Any institution would be extraordinarily fortunate to have him."
"I am confident he will consistently demonstrate unwavering commitment and diligence in all his endeavors. I extend my best wishes to him for his future pursuits."
Gallery
