Ai face guard
App detects unauthorized photo use via face recognition. Alerts user via SMS/app to Aadhaar-linked phone. User taps ALLOW to permit or DENY to block/delete upload instantly.
Updates
For AI Face Guard, I took ownership of the real-time face detection pipeline, from hardware capture to alert delivery. My job was to make sure the camera’s live feed could be processed by our AI model quickly and accurately.
I started by wiring and configuring the camera hardware, then integrated it with the AI inference engine using a face recognition API. Getting the two systems to talk smoothly was far from trivial — frame drops, mismatched formats, and latency spikes were constant hurdles.
To fix this, I implemented frame buffering and async processing, which allowed the AI to analyze video frames without blocking the capture stream. I also tuned the image resolution dynamically so the model ran efficiently on our limited hardware without losing detection accuracy.
Once the detection worked reliably, I built the alerting system so that whenever a flagged face was spotted, the system would instantly send notifications to the dashboard (and optionally to a mobile device). This required webhook integration and API rate optimization so alerts were fast but never spammy.
What I’m most proud of is that by the end, AI Face Guard could identify and log faces in under 200ms, even over a low-bandwidth network. It felt like building a security guard that never sleeps.
I learned more in this build than I expected — from edge computing optimization and real-time video streaming to API debugging and system resilience under unreliable connections. The whole experience taught me how to bridge the gap between AI theory and hardware reality.
Leave feedback in the comments!
Log in or sign up for Devpost to join the conversation.
Log in or sign up for Devpost to join the conversation.