A YOLOv8-powered real-time safety monitoring system for railway stations
A real-time object detection system designed to identify humans, emergency gestures, and crime alerts in public places like railway stations using a custom-trained YOLOv8 model.
Railway stations in India face security challenges due to growing passenger volumes and limited real-time monitoring. Traditional CCTV surveillance lacks automation and quick response capabilities. This project addresses this gap by building an AI-powered alerting system using sign gesture recognition.
We implemented a lightweight YOLOv8-based solution that detects humans and hand gestures corresponding to emergencies or crimes in real-time video feeds. The system is trained on a custom dataset and provides alerts using audio and visual cues for quick response.
- 🎯 Real-time detection of:
- Human presence
- Emergency gestures
- Crime-indicating gestures (e.g., 'X' sign from sign language)
- 📦 YOLOv8 custom model trained on Roboflow-annotated dataset
- 🔊 Audio alert system for real-time response
- 💻 Runs in Google Colab (training) and local machine (inference)
- 📊 Evaluation metrics: Precision, Recall, F1-score, mAP@0.5, mAP@0.5:0.95
- Python 3.10
- Ultralytics YOLOv8
- OpenCV
- Google Colab (for training)
- Flask (optional for deployment)
- Roboflow (for dataset creation and augmentation)
