Self-driving Public Mobility Get-off Safety System
deep-learning
lstm
computer-vision
yolo
Bus passenger get-off intention prediction using YOLOv5 + DeepSort + LSTM
1 Overview
AI system for predicting passenger get-off intentions to ensure safe operation of autonomous buses.
2 Tech Stack
| Category | Technology |
|---|---|
| Object Detection | YOLOv5 |
| Object Tracking | DeepSort |
| Pose Estimation | Multi-person Skeleton Detection |
| Sequence Model | LSTM |
| Framework | TensorFlow / Keras |
| Language | Python |
3 Key Features
- Real-time Passenger Tracking: Detect and track passengers inside the bus using YOLOv5 + DeepSort
- Skeleton Data Extraction: Convert each passenger’s skeleton data into time-series format
- Get-off Intention Prediction: Binary classification (get-off/stay) using LSTM analysis of data 35 frames before departure
- High Accuracy: Achieved stable convergence and high classification accuracy on validation data
4 My Role
5
6 Team
- Team: Arrival