Pingpong Ball Detection

1. Goal of this project

The goal of this project is to detect pingpong ball with YOLOv3.

2. Developed Environment

OS : ubuntu18.04 ROS melodic Processor : Nvidia TX2 Nvidia Kernel : Jetson TX2 Kernel
This program is for Xycar(RC-Car). So a "weights(model)" file is converted to tensorrt sutable for architecture of Xycar. So if you want to lighten your model to fit your hardware, you should make your own tensorrt.
If you don't want, just use the model(a weights file).

pips required for this project are written on "requirements.txt"

$ pip install -r requirements.txt

3. Main Scripts

1. trt_detection.py

This script is to subscribe detected objects and control next activation of Xycar. In this case, it prints the color of pingpong ball out.

2. trt_distance.py

This script is to compute distance of detected pinpong ball. It is calculated based on Geometrical Distance Estimation.

3. trt_inference.py

This script is to inference object if you convert a weights file to tensorrt.

4. How to Run

below is the example code.

$ roslaunch detection detection.launch