NVIDIA Raccoon Sentry Bot
This video demonstrates the Racoon bot detecting and deterring raccoons in a real-world setting.
Key Learnings
- Real-time object detection on edge device
- Combining computer vision with robotics
- Importance of robust testing in real-world conditions
Project Overview
I had a problem where raccoons lived near my house and kept getting into my trash bins. This inspired me to develop a sentry bot using NVIDIA Jetson hardware to create an automated pest control solution. This project combined computer vision with robotics to squirt raccoons with water to deter them from my trash bins. Project scope included:
- Implemented real-time object detection using SSD MobileNet architecture
- Fine-tuned model on OpenImages dataset specifically for raccoon detection
- Designed and integrated squirt gun with servo motor to spray water
Technical Details
- Collected and cleaned images with bounding boxes of raccoons from OpenImages dataset
- Fine-tuned SSD MobileNet model using NVIDIA's libraries
- Implemented custom detection threshold logic to minimize false positives
- Developed system for water spray mechanism using GPIO
- Logged detections to a file to analyze later
Technologies Used
NVIDIA Jetson Nano
TensorRT
SSD MobileNet
Computer Vision
GPIO
Python
OpenCV