README_en.md

Based on Yolov5 + onnx + PyQt5 + nuitka

---

English | 简体中文

*If you encounter an unsolvable problem, issue are allowed*

---

• Implementation
• Preview
  • download the demo
  • main window
  • feature
• Requirement
• Tutorials
  • quickstart
  • train custom data and export to onnx model
  • packaging an executable file

---

Implementation

• Train custom dataset with yolov5
• Export pt format to onnx format
• Inference with openCV.dnn or onnxruntime
• Package as executable program on Windows platform (Linux can also be packaged in theory, but it has not been tried)
• Portable (deployed) to most Windows devices after packaging

---

Preview

download the demo：https://download.kstore.space/download/3190/plugin/publish.zip (130MB)

main window:

Feature

1. support image, video, webcam, RTSP/RTMP/HTTP streams, screenshot
2. real time frame rate
3. redirect stdout to GUI
4. change conf_thres and iou_thres at any time
5. display/no bounding box, change bounding box's color
6. print/hide inference result
7. record video
8. save screenshot or log
9. custom script, triggered on each picture(see the need/self_demo.py)

---

Requirement

• python >= 3.9
• numpy == 1.23.4
• opencv-python == 4.5.5.62
• PyQt5 == 5.15.7
• onnxruntime == 1.13.1
• nuitka == 0.6.18.4

---

Tutorials

Quickstart

• Clone this Repo
• Install requirements: pip install -r requirements.txt
• Run Yolo2onnxDetectProjectDemo.py
• Then you will see the GUI
• Click ▶ button

Train custom data and export to onnx model

• Yolov5 v5.0 is recommended
• How to train?目标检测---教你利用yolov5训练自己的目标检测模型
• Export to onnx format: python models/export.py --weights ./weights/yolov5s.pt --img 640 --batch 1

Packaging an executable file

• The module used is nuitka, and the command has been configured in build.py. If you need more advanced playing methods, please see https://nuitka.net/
• Run build.py, finished products are located in build_file/publish folder
  1. Tips:the truly finished products are in the Yolo2onnxDetectProjectDemo.dist folder
  2. To facilitate debugging and updating, all files in this folder need to be copied to the publish folder after the first packaging is successful
  3. Double click to run the exe file, and copy the module to the publish folder according to the exception, until it runs successfully

---