YOLOv5-tensorflow-lite-Raspberry-Pi

This GitHub repository show real-time object detection using a Raspberry Pi, YOLOv5 TensorFlow Lite model, LED indicators, and an LCD display.

This GitHub repository show real-time object detection using a Raspberry Pi, YOLOv5 with TensorFlow Lite framework, LED indicators, and an LCD display. the feature of this project include:

• Show fps for each detection
• Output the class using LED for each class (there is 5 classes: car, person, truck, bus, motorbike)
• Show CPU and temperature of raspberry pi using LCD 16x02.

Demo

Below is the following demo video showcasing the Raspberry Pi in action. When real-time object detection processed, video frames show the fps, LED indicators will trun on based on detected classes, and CPU usage and temperature information displayed on the LCD screen.

mAP and FPS

Model	mAP@50	mAP@50:5:95	FPS
Yolov5s 640px fp32	94,7	74,1	0,5
Yolov5s 416px fp32	91,8	72,5	1,1
Yolov5s 320px fp32	90,5	69,8	1,87
Yolov5n 640px fp32	91,4	67,3	1,5
Yolov5n 416px fp32	89	66,3	3,7
Yolov5n 320px fp32	86,7	63,7	5,7
Yolov5s 640px int-8	93,9	70,4	0,7
Yolov5s 416px int-8	90,5	67,5	1,7
Yolov5s 320px int-8	90,1	63,9	2,9
Yolov5n 640px int-8	90,7	64,4	1,9
Yolov5n 416px int-8	88,7	63,2	4,5
Yolov5n 320px int-8	85,9	59,3	7,2

Prerequisites

• Raspberry Pi 4 (I'm using 8 GB version)
• Raspberry Pi OS 11 Bulleyes 64-bit
• Pi Camera v2/v1/Web-Camera
• PCB or PCB Dot
• LCD 16x2 Biru/Blue 1602 SPI I2C
• ✨ Wiring cable ✨

Wiring Diagram

Follow this organized table to establish the proper connections, you can also read the reference here GPIO on Raspberry Pi4.

LED Wiring - Raspberry Pi

Wire Color	GPIO Pin
Red	GPIO 17
Green	GPIO 18
Yellow	GPIO 23
Cyan	GPIO 27
White	GPIO 22
Black (GND)	GND

I2C Wiring - Raspberry Pi

Wire Color	Connection
Red	5V
Black	GND
Purple	SDA
Brown	SCL

Installation

To run this project, you need Python 3.5 or higher installed on your system. Follow these steps to get started:

• Clone the repository and navigate to the project directory: :

  git clone https://github.com/kiena-dev/YOLOv5-tensorflow-lite-Raspberry-Pi.git
  cd YOLOv5-tensorflow-lite-Raspberry-Pi

• Create a Python virtual environment (optional but recommended):

  python3 -m venv venv

• Activate the virtual environment:

  source venv/bin/activate

• Install the required dependencies using pip3:

  pip3 install -r requirements.txt

Now you have successfully installed the project and its dependencies.

Usage

$ python detect.py --weights yolov5s.pt --source 0                                   # webcam
                                                     img.jpg                         # image
                                                     vid.mp4                         # video
                                                     path/                           # directory
                                                     'path/*.jpg'                    # glob
                                                     'https://youtu.be/Zgi9g1ksQHc'  # YouTube
                                                     'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream

example video

Default (without LED/LCD):

  python detect.py --img 320 --weights yolov5n_320px-fp32.tflite --source video_test.mp4  

With LED/LCD:

  python detect_rpi_led.py --img 320 --weights yolov5n_320px-fp32.tflite --source video_test.mp4  

example webcam

Default (without LED/LCD):

  python detect.py --img 320 --weights yolov5n_320px-fp32.tflite --source 0

With LED/LCD:

  python detect_rpi_led.py --img 320 --weights yolov5n_320px-fp32.tflite --source 0

Training Dataset

If you want to train your own model, you can utilize the resource provided below:

Dataset from Roboflow:

Be sure to make use of these resources to train your model and achieve optimal results!

Tips

You can change your own class, add or modify in coco128.yaml. modify the code below:

names:
- bus
- mobil
- honda
- orang
- truck
nc: 5
roboflow:
  license: CC BY 4.0
  project: skripsi-dtmyf
  url: https://universe.roboflow.com/devan-naratama-2xq45/skripsi-dtmyf/dataset/2
  version: 2
  workspace: devan-naratama-2xq45
test: ../test/images
train: /devan/datasets/Skripsi-2/train/images
val: /devan/datasets/Skripsi-2/valid/images

you can change your own class!

Authors

• @kiena

Reference

Special thanks to the following resources that inspired and contributed to this project:

• YOLOv5 Github Page as main source
• QEngineering
• Tensorflow