Surveillance.AI

Inspiration

Our project, Surveillance.AI, was born out of a common frustration faced by our friend who struggled with extracting valuable insights from lengthy video footage. We realized the need for a solution that could efficiently analyze video data, saving time and effort in the process.

What it does

Surveillance.AI leverages AI technology to sift through extensive video content and extract relevant data, which is then stored in a knowledgebase. When users input prompts containing specific search criteria, the AI identifies the relevant segments within the video, retrieves timestamps, and trims the footage accordingly, providing precise results.

How we built it

We crafted Surveillance.AI using the following tech stack: Python, YoloV8, Langchain, Prisma, PostgreSQL, and FastAPI for the backend, along with NextJSand TypeScript for the front end. This combination enabled us to develop a robust system capable of seamless video analysis.

Challenges we ran into

Several challenges emerged during the development process. Creating a comprehensive dataset for our use cases proved time-consuming, and hardware limitations without a GPU caused processing delays. Additionally, training the Language Learning Model (LLM) to accurately process user prompts posed its own set of challenges.

Accomplishments that we're proud of

Despite the hurdles, we're proud to have developed Surveillance.AI within a short timeframe. Overcoming technical challenges and successfully implementing a functional system underscores our team's dedication and expertise.

What we learned

Our journey with Surveillance.AI taught us invaluable lessons. We delved into Langchain and its Kor feature, which facilitated structured data extraction using LLMs. Furthermore, we deepened our understanding of Yolo models for object detection and honed our skills in Next.JS frontend development

What's next for Surveillance.AI

Looking ahead, we envision expanding Surveillance.AI to encompass a broader range of use cases beyond person and vehicle detection. Future iterations could include applications in crowd management, litter detection, fight detection, theft prevention, and more. Our goal is to continue refining and enhancing the system to address evolving surveillance needs effectively.