INVESTigator

A Community-Driven Investigation Platform

❮

❯

INVESTigator was inspired by the common missing person posters we see in our communities. This platform aims to reduce the mental load on detectives by leveraging community contributions to ongoing investigations. Witnesses can upload evidence, which is automatically filtered for relevance while preserving all data. The system provides real-time data insights using websockets and multiprocessing, and includes features like license plate detection, similarity analysis with detective databases, and audio analysis. We built it using Django for the frontend, with HTML, CSS, and JS, along with an sqlite3 database. For a smooth user experience, we integrated websockets and AJAX. The backend leverages PyTorch, TensorFlow, and OpenCV for data analysis. We used an LSTM model for sentiment analysis and a CNN model for facial emotional analysis, combining sequential and spatial data handling for comprehensive analysis. Despite challenges with integrating models and creating a visually appealing interface, we are proud of our finished product. It’s an ambitious project that we’re excited to share. I learned valuable skills in websockets, AJAX, and video analysis with ML. Looking ahead, I may develop more accurate models, add features for holistic analysis, and enhance community engagement with an explore page and data summarization.