Real-Time Weapon Detection AI

I need a computer-vision solution that detects harmful weapons—specifically guns and knives—in live video and responds instantly. The moment the model recognises one of these threats, it must do two things without delay: • trigger an audible alarm on-site, and • send real-time alerts (image or short clip plus timestamp) to designated security personnel via a simple dashboard, mobile push, or email/SMS gateway. A live view overlay that highlights the detected weapon would be ideal so guards can follow the situation on existing CCTV monitors, but the core requirement is the alarm and secure alert pipeline above. You are free to choose the most suitable stack—Python, TensorFlow/PyTorch, YOLOv8 or equivalent—provided the final system runs reliably on an NVIDIA Jetson or modest GPU server and can be integrated with RTSP/IP cameras. I will supply sample footage for initial testing; please allow for additional fine-tuning with my data once the first working model is up. Deliverables 1. Trained detection model with at least 90 % precision on guns and knives in varied lighting. 2. Source code (inference script, alert logic, config files) and a brief README for deployment. 3. Dockerfile or install script so I can replicate the environment. 4. Short video demo proving the alarm and alert flow in real time. Acceptance is based on a live test over my camera feed where detection, alarm and alert reach the security console in under three seconds.

Project ID: 39730623