pru-robotic / jetson-rpi-image-processing

Base Image Processing repo focuses on YOLO (You Only Look Once) used in Jetson Nano (+ Orin Nano) & Raspberry Pi 4

Image Processing | Object Detection

• Purpose
• Features
  • Prerequisites
  • Notes
• Project Structure
• Raspberry Pi
  • How To Run?
• Jetson Nano
  • Jetson Orin Nano
• Licence

Purpose

Base Image Processing and custom Object Detection repo focuses on YOLO (You Only Look Once) used in Jetson Nano (+ Orin Nano) & Raspberry Pi 4.

It also contains various real time image processing applications and related with electrical components such as Pixhawk and Servo motors.

Main purpose of this repo is creating an image processing software base in mainboards for practicing and adapting the current experiences for the next generation of our technical project members.

Features

Prerequisites

• Raspberry Pi 4 - 4GB (at least)
• Jetson Nano 4GB
• Jetson Orin Nano
• USB Camera
• SD Card (32GB at least) and reader

Notes

1. Install an OS for Raspberry and Jetson then use: !Warning: Need to install the 64-bit OS for Raspberry Pi due to YOLO's ultralytics library configurations in python.

2. Using USB Camera for higher FPS and better software quality rather than CSI cameras. Further, 64-bit OS is not suitable with CSI cameras in general.

3. For USB cameras opencv index is: In Raspberry Pi 4, videoCapture(1) and Jetson Nano devices, videoCapture(0).

4. WIP: tensorflow library is unavailable to install and use in Raspberry Pi 64-bit OS, currently. Click to see the issue!

Project Structure

The project follows this directory structure:

jetson-rpi-image-processing/
│
├── cfg/
│   ├── yolov3_testing.cfg
│   └── yolov4-tiny-custom.cfg
│
├── components/
│   ├── gpio_guide.py
│   ├── mav_pixhawk.py
│   ├── mavlink_object-detect.py
│   ├── mesafe_sensor.py
│   └── servo.py
│
├── object_detection/
│   ├── coco.names
│   ├── object_detection.py
│   ├── realtime_core.py
│   ├── realtime_gpio.py
│   ├── realtime_ultralytics.py
│   └── realtime_v1.py
│
├── test_images/
│   └── ...
│
├── training/
│   └── train_YoloV3.ipynb
│
├── weights/
│   ├── best.pt
│   ├── yolov8n.pt
│   ├── yolov3_training_last.weights
│   └── yolov4-tiny-custom_last.weights
│
├── gitignore
├── README.md
└── requirements.txt

• cfg/: Contains config files for yolov3 and yolov4-tiny.
• components/: Contains code samples related with electrical components such as Pixhawk and Servo motors.
• object_detection/: Contains image processing and object detection codes in both real time and normal.
• test_images/: PNG or JPG image files for testing.
• training/: Google Colab yolov3 training notebook.
• weights/: Custom trained weight files for yolov3, yolov4-tiny and yolov8
• requirements.txt: Lists project dependencies.

Raspberry Pi

How To Run?

1. Virtual environment setup:

python3 -m venv yolovenv

2. To activate the virtual environment (Windows):

yolovenv/Scripts/activate

3. To activate the virtual environment (Linux / MacOS):

source yolovenv/bin/activate

4. Install dependencies:

• If you are running repository on hardware platforms other than Raspberry Pi or Jetson, you should remove RPi.GPIO library from requirements.txt file; because this library is designed specifically for devices equipped with GPIO pins and may not be compatible with other platforms.
• The ultralytics library may not be compatible with all Python environments depending on your device but it's essential for certain features, particularly on Raspberry Pi with 64-bit OS.

pip install -r requirements.txt

or

pip3 install -r requirements.txt

5. Run:

python file_name.py

or

python3 file_name.py

Jetson Nano

Jetson Orin Nano

For Jetson Nano (4GB) and Jetson Orin Nano (8GB), there is no need to use python environment for running ultralytics library as an extra. It can be installed and used, directly:

pip install ultralytics

or

pip3 install ultralytics

Licence

This project is licensed under the MIT License - see the LICENSE file for details.

jetson-rpi-image-processing's Issues

Tensorflow cannot install in Raspberry Pi 4 OS with 64-bit

WIP (work in progress):

• Raspberry Pi's current OS is the Bullseye 64-bit due to compilable with ultralytics which is a YOLOv5 and newer versions library.
• tensorflow cannot be installed in Jetson Nano, too :(