We develop real-time computer vision models, neural face detection, object tracking, and edge AI pipelines that bring visual intelligence to camera systems and autonomous hardware.
From high-speed video feeds to embedded edge hardware, we build low-latency computer vision pipelines.
Real-time face detection, object localization, and YOLO-based boundary classification on live video streams.
Biometric facial landmark extraction, identity verification, and visual feature classification for access and telemetry.
Continuous click-to-lock target identification, centroid trajectory calculation, and spatial displacement tracking.
Embedded vision algorithms optimized for edge processors, robotic gimbals, and autonomous drone flight loops.
A real-world applied computer vision and drone flight control project developed in Python and OpenCV with DJI Tello integration. Features real-time neural face detection, click-to-lock target selection, and autonomous target following.
Expensive computer vision inference originally bottlenecked the 50Hz flight control loop. We re-architected the system to decouple optical processing from flight stabilization using asynchronous worker threads.