Technical Implementation & AI Facial Recognition Accuracy: "I see you're using a face tracking system in camera-backend. How does the system handle different lighting conditions, occlusions (like masks or glasses), or multiple people in the frame at once?" Copilot API Integration: "One of your recent commits mentions integrating the Copilot API for suspicious activity detection. How exactly are you defining 'suspicious activity' in your prompts, and how do you handle potential false positives?" Backend Scaling: "The project uses Python for video processing and Vue for the frontend. How would this system scale if you needed to monitor 50 cameras simultaneously instead of just one?" Security & Privacy Data Privacy: "Security systems naturally capture sensitive biometric data. How is the facial data stored, and what measures have you taken to ensure that the logs and face crops in camera-backend are encrypted or protected from unauthorized access?" Authentication: "You recently added Auth0 Authentication. Is this used only for the dashboard, or is it also integrated into the API endpoints that handle the live camera feeds?" Edge vs. Cloud: "Does the video processing happen locally on the 'edge' device, or is the stream sent to a server? What are the latency implications of your chosen architecture?" User Experience & Features Real-time Alerts: "How does the 'Muninn' system notify a user when an anomaly is detected? Is there a push notification system, or does the user have to be actively looking at the hackathon-frontend?" The 'People' Modal: "The start.ps1 file mentions a People modal UI and face-crop API. Is this intended for 'enrolling' known users (like family members) to distinguish them from strangers?" Hardware Requirements: "What kind of hardware is required to run the camera-backend efficiently? Could this run on a low-power device like a Raspberry Pi, or does it require a dedicated GPU?"
OUR Pitch
INTRO Hi, everyone, Rohit, Aditya, Gabriel, and Lokesh, so what if your house didn't just record a crime, but recognized the intent? For decades a lot of security companies have been asking the question “What happened?”. With Tiger Sentinel we’re moving security to answer, “What is about to happen”.
PROBLEM For the majority of security systems, whether it’s a residential home, a small business or a local bank, it’s a retroactive record. A lot of these systems that are used for day-to-day lack key facial recognition technology. When standard security measures fail, owners have no way they are under attack until the damage is already done. This application is proactive, affordable, and widely accessible to the public.
Solution The reasons above are our main reasons why we created Tiger Security This is a real time facial detection to identify individuals the moment they are able to get into frame of the camera. This just doesn’t record, it actively cross-references faces with a database. If a known threat is picked up by the system, it would immediately recognize it and give owners a crucial head start before the crime is committed.
Technology and Demo
The Backend We used Python and OpenCV to handle the heavy lifting of continuous video stream analysis. We also optimized the detection models to pick up edge cases such as a side profile and having multiple individuals within the frame of the camera without crashing under load
API - The core of our data pipeline relies on an API that connects our backend infrastructure to the frontend dashboard. Instead of having raw data, the API pushes the live video feed and the facial recognition database straight from the backend to the user’s screen in real time.
Database When the system flags an unknown person, it would dynamically automate and take a snapshot of the person, then the owner could go in and update the profile. If the owner knows the person he could deem them as safe, if not they could deem the person as a risk.
Frontend We tied all of this together with a clean, responsive frontend dashboard built using the Vue.js framework. It’s secured by Auth0, so the owner could easily track their alerts
Log in or sign up for Devpost to join the conversation.