Pothole Detection and Road Damage Reporting System
Developed a deep learning system to detect and classify road potholes from images using the YOLOv8 object detection model. Built a REST API using FastAPI to process uploaded images and return identified pothole locations and severity levels. Designed a React-based web interface allowing users to upload road images and visualize detected potholes. Integrated geolocation-based reporting so detected potholes can be mapped and prioritized for municipal repair. Implemented a dashboard to visualize pothole reports and metadata for monitoring road conditions.
Agrithon: Crop Yield Prediction System
Built a regression-based machine learning model to predict crop yield using environmental and soil parameters. Developed a preprocessing pipeline using Scikit-Learn for categorical encoding, feature scaling, and model inference. Created an interactive Streamlit dashboard enabling users to input parameters and obtain real-time yield predictions.
Image Compressor Web Extension
Created a Chrome extension (Manifest V3) that captures the current browser tab and allows users to compress images directly. Built a local FastAPI backend using Pillow to compress images with configurable format (JPEG/WebP), quality, and resizing options. Implemented a browser-based fallback compression mechanism using the Canvas API when the backend server is unavailable.
Footprint: Gamified Carbon Tracking Platform
Architected a full-stack web application to track, visualize, and gamify personal CO2 emissions. Made a custom user authentication system with personalized data storage to manage individual user profiles. Engineered the core tracking and analytics engine to provide real-time visual insights into carbon usage. Designed a gamification module that incentivizes eco-friendly user behavior through measurable reward metrics.