K-ompanion: Smart Device for Dog Training
deep-learning
yolo
edge-computing
jetson
Dog behavior recognition and interaction robot using YOLOv5 + Jetson Nano
1 Overview
Intelligent robot device that recognizes dog behaviors (bodylower, feetup, footup, sit, etc.) and interacts accordingly. Submitted to AI Robot Hackathon Magok 2021.
2 Tech Stack
| Category | Technology |
|---|---|
| Object Detection | YOLOv5 |
| Edge Device | Jetson Nano |
| Model Optimization | TensorRT |
| Dataset | AI Hub Dog Behavior Recognition Dataset |
| Framework | PyTorch |
| Language | Python |
3 Key Features
- Dog Behavior Recognition: Real-time detection of specific poses (bodylower, feetup, footup, sit)
- Interactive Response: Performs activities like ball throwing in response to recognized behaviors
- Edge Optimization: TensorRT conversion for real-time inference on Jetson Nano
- Model Experiments: Validated alternative models including RetinaNet and EfficientNet alongside YOLOv5
4 My Role
5
6 Event
- AI Robot Hackathon Magok 2021