Wheelly the Coyote - Object Detection on Raspberry Pi 4B / 3B+

End-to-end object detection using Raspberry Pi 4B / 3B+

Table of Contents

Install
Usage
File Details
Bill Of Materials
Troubleshooting

Install

Software

1. Clone this repository

git clone https://github.com/RootLeo00/robot-wheelly-object-detection.git

2. Download requirements

cd robot-wheelly-object-detection
chmod +x setup.sh
./setup.sh

Usage

3. Start pigpiod daemon

sudo pigpiod

4. Start detection

python detect.py

File details

├──  test  
│    └── testservo.py  - a test to understand if servo is working
│    └── testsonar.py  - a test to understand if sonar is working
│    └── testwheels.py  - a test to understand if motors, wheels and h-bridge are working
|
|
├──  detect.py - file with main that runs the program  
├──  efficientdet_lite0.tflite - pretrained model using EfficientDet0 neural network
├──  efficientdet_lite0_edgetpu.tflite - pretrained model using EfficientDet0 neural network to use with EdgeTPU
├──  requirements.txt  - file with the names of the python modules to download
├──  servo.py - class with Servo functions
├──  setup.sh - file to run to download tflite models and requirements
├──  sonar.py - class with Sonar functions
├──  utils.py - utilities, such as function visualize
├──  wheels.py - class with functions to control motors and wheels

Bill Of Materials

• Raspberry PI 4B
• 2 Trolley Chassis
• 4 Micro DC Motor (Geared) - 90 RPM (6-12V)
• 4 wheels
• L298N Dual H-Bridge Motor Controller
• Ultrasonic Sensor - HC-SR04 (Sonar)
• Jumper wires MM / FM / FF
• USB-C wire
• Cooling Fan
• Power Bank 5200mAh
• 4 batteries AA (1.5 V)
• Raspberry Pi Camera Module 1.3
• Micro Servo 9G

Troubleshooting

1. Camera not found

• Make sure that you have enabled camera interface on Raspberry Pi, through command:

sudo raspi-config

• Make sure your system is detecting camera. Try running:

  vcgencmd get_camera

  You should get back supported=1 detected=1, indicating that the camera is detected and supported by the operating system. If you get detected=0, then the camera is not being seen by the operating system. You can try to force OS to auto detect camera by modyfing boot configuration:

  sudo vim /boot/config.txt

  Write somewhere on file:

  camera_auto_detect=1

  Restart system

  reboot

• detect.py use deafult video device on /dev/video0, if you want to change to another (e.g. /dev/video1), you have to launch the program passing the --cameraId argument. Example:

python detect.py --cameraId=1

2. Raspberry is slow on detect the object

   This is totally normal. If you want to have a faster and more accurate detection, you can use EdgeTPU (e.g. Coral Accelerator) or another type of board

3. Wheels, sonar or servo don't work

   Try to run their specific tests. You can find the tests in /test folder.