Contact person: AI Railway Safety Alert System

Phone:Show

Email:Show

Location: Kochi, India

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I want to build an end-to-end, vision-based safety platform for railway environments that can reliably spot three critical events in real time—intrusion, obstacle presence on the track, and trespassing—and immediately trigger an audio alarm both on-site and at a central control room.

The detection engine must rely on a YOLO-based model (YOLOv5, v8 or equivalent) trained or fine-tuned for railway scenarios. I already refer to it as “YOLO-UAT,” so the delivered code and weights should follow that naming and include clear documentation of the training pipeline, datasets used, and performance metrics. False positives have to be kept low enough that normal operations are not disturbed, yet reaction time from camera frame to alarm activation should remain under one second.

Audio signalling needs to fire simultaneously at the trackside device (for field workers or trespassers to hear) and at the central monitoring desk (for operators). I’m open to whether you choose Raspberry Pi, Jetson, Arduino-driven relays, or another embedded controller for the alarm relay, as long as the final design is fully explained and reproducible.

Deliverables
• Annotated source code for detection and alarm logic (Python preferred)
• Trained YOLO-UAT weights with at least 90 % mAP on the supplied validation set
• Hardware wiring diagram and BOM for the dual-location audio alarm
• Step-by-step deployment guide covering setup on edge device, camera integration, and remote audio link
• Short video or live demo link showing the system flagging each of the three target events and sounding alarms at both locations

Acceptance will be based on a live test or recorded video that proves reliable detection of intrusion, obstacle, and trespassing scenarios, followed by synchronized audio alerts at the track and the control room." (client-provided description)