Project Highlights
3D Percecption · Pytorch, MMCV
S2TPVFormer: A Spatiotemporal Transformer for 3D Semantic Occupancy Prediction
Accurate and comprehensive 3D scene understanding and reasoning are pivotal for the development of robotic and autonomous driving systems. The study focuses on advancing 3D scene understanding in the scope of 3D Semantic Occupancy Prediction, emphasizing spatial and temporal reasoning as key components. Leveraging TPV (Tri-Perspective View) representation, our spatiotemporal encoder generates temporally rich embeddings, fostering coherent occupancy predictions. The study proposes a novel Temporal Cross-View Hybrid Attention (TCVHA) mechanism, enabling the exchange of spatiotemporal information across TPV views.
Embedded Systems · C++
forté: The Remote Keyboard Tutoring System
forté is a web-based system that can be attached to any (electronic) keyboard synthesizer through a MIDI connector. Once our system is connected to the keyboard, the user can interactively learn, play or teach in combination with the web application that we provide.
Machine Learning · Python, Scikit-learn, TensorFlow
Sobriety Detection using Mobile Phone Gyroscope Data
An intelligent API which is capable of passively tracking gyroscopic data to classify sobriety level in real-time that will help minimize risks of drunk-riding accidents for e-scooter sharing systems.
Natural Language Processing · Python, TensorFlow, Keras, NLTK
Conversational Transformer Chatbot
This project was done as part of the final honors assignment in Natural Language Processing course on Coursera. In this project, I've built a conversational chatbot that can understand and respond to natural language questions.
Image Processing · Python, OpenCV
A License Plate Recognition (NPR) System
This project was done as an assignment in the course "Image Processing (CO543)" offered by the Department of Computer Engineering, University of Peradeniya. The main objective of this project was to develop a license plate recognition system that can detect and recognize license plates in images. The system was developed using OpenCV and Python.