CCTV–Based Abnormal Event Detection System

An AI-driven surveillance platform that automatically detects unseen abnormal or suspicious human activities from live CCTV footage in real-time using deep learning models. The system can identify anomalies it has never been explicitly trained on, enabling zero-shot detection of novel threats and suspicious behaviors. This supports proactive threat detection and automated authority notifications to improve public and institutional safety.

GitHub: View Repository

Key Features:

• Unseen anomaly detection — identifies novel, previously unknown abnormal behaviors without explicit training
• Real-time anomaly detection from CCTV streams
• Live dashboard with video preview and prediction overlays
• Deep learning pipeline (EfficientNet + BiLSTM + Transformer)
• Interactive user feedback system for active model learning
• Automated authority notification (Email/SMS via Twilio & SendGrid)
• Scalable microservice backend built with Flask API
• Feedback-based incremental model fine-tuning
• Detection logs with timestamps and confidence scores
• Deployed for intelligent city surveillance and research use cases

Technologies Used:
Python, Flask, PyTorch, React.js, OpenCV, Twilio, SendGrid, Firebase, Tailwind CSS, RESTful APIs

Research Domain:
Computer Vision, Deep Learning, Video Surveillance, Unseen anomaly detection

Vehicle & Transportation Reservation System (Vishwa Transport)

A comprehensive web-based platform enabling customers to easily book vehicles while providing administrators full control over fleet, schedules, and reservations.

GitHub: View Repository

Key Features:

• Online vehicle booking & scheduling
• Vehicle CRUD (Create / Read / Update / Delete) with images & specs
• Staff & driver details management (roles, contact, assignments)
• Booking list & booking history with filtering and export
• User & Admin dashboards with role-based access
• Vehicle availability tracking & automated conflict detection
• Secure authentication (JWT) and REST API
• Responsive, modern UI — mobile & desktop friendly

Tech Stack:
React, Node.js, Express, MongoDB, JWT Auth, REST API

Dairy Food Store Room Management System

A full-featured inventory & store room management system for dairy products — enabling staff management, food item tracking, order handling, and automated stock monitoring to ensure product freshness and operational efficiency.

GitHub: View Repository

Tech Stack:
React, Node.js, Express, MongoDB, JWT Auth, REST API, Bootstrap

Personalized Weather Application (for Mr. Mickey Arthur)

A personalized web-based weather application that allows authorized users to log in and access real-time weather information. The system provides current weather in Colombo, location-based search, and both short-term and weekly forecasts with a clean and responsive UI.

GitHub: View Repository

Tech Stack:
React, Node.js/Express (for auth), OpenWeatherMap API, LocalStorage/Session, CSS/Bootstrap

AI-Powered Hand Sketch Recognition & Colorization System

An advanced AI system that recognizes hand-drawn sketches and transforms them into realistic colored images using deep learning, vision transformers, and diffusion models. Users can draw sketches or upload existing ones, and the system predicts the object while generating high-quality photo-realistic outputs.

GitHub: View Repository

Tech Stack:
PyTorch, HuggingFace Transformers, Stable Diffusion, ControlNet, OpenCV, Streamlit, Python

Models & Pipelines:
BEiT-Sketch Classifier, Stable Diffusion v1.5, ControlNet-Scribble

Hybrid AI-Based Fake News Detection System

An AI-powered fake news detection system that combines machine learning with semantic fact-verification. The model analyzes news articles, identifies key entities, checks facts against trusted knowledge sources like Wikipedia, and determines whether content is credible or misleading.

Tech Stack:
Python, scikit-learn, spaCy, HuggingFace SentenceTransformer (MiniLM), Pandas, Flask API, joblib

Future Enhancements:
Transformer-based classification (BERT/GPT), multilingual support, cloud deployment & MLOps pipeline

AI-Powered Customer Support Chatbot

An intelligent conversational AI system designed to automate customer support operations with natural language processing, intent recognition, and context-aware responses. The chatbot provides instant, accurate answers to customer queries while seamlessly escalating complex issues to human agents.

GitHub: View Repository

Technologies Used:
Python, Flask/FastAPI, OpenAI GPT API / Hugging Face Transformers, React.js, Socket.io, MongoDB, Redis (session management), NLTK/spaCy, RESTful APIs

Application Domain:
Natural Language Processing (NLP), Conversational AI, Customer Service Automation, Chatbot Development

Student Marks Analyzer

A comprehensive web-based analytics platform for student performance evaluation built entirely in Haskell. The system provides real-time statistical analysis, interactive data visualization, and automated report generation, modeling ETL-style data pipelines commonly found in education analytics and institutional reporting systems.

GitHub: View Repository

Technologies Used:
Haskell (Backend), Scotty Web Framework, Aeson (JSON parsing), HTML5, CSS3, Vanilla JavaScript, RESTful API, CSV Processing, Functional Programming Paradigms

Application Domain:
Education Analytics, Data Processing, Academic Performance Evaluation, ETL Data Pipelines, Functional Programming

Team Members:
Sewwandi L.L.C, Shamali A.V, Welikumbura R.W.R.L.