English | 简体中文

训练策略

This system is a road sign recognition application leveraging YOLOv5🚀 😊. It employs a MySQL database 💽, PyQt5 for the interface design 🎨, PyTorch deep learning framework⚡. Additionally, it incorporates CSS styles 🌈.

The system comprises five key modules:

1. System Login Module 🔑: Responsible for user authentication.
2. Initialization Parameter Module 📋: Provides settings for initializing YOLOv5 model parameters.
3. Sign Recognition Module 🔍: The core functionality responsible for recognizing road signs and updating the database with the results.
4. Database Module 💾: Consists of two sub-modules - basic database operations and data analysis.
5. Image Processing Module 🖼️: Handles the processing of individual images and associated data.

The entire system is designed to support various data input methods and model switching. Additionally, it offers image enhancement techniques such as mosaic and mixup 📈.

• Sign Recognition Module

The three checkboxes in the lower left corner are results save, start database entry, and model visual analysis.

• Image Processing and Data Augmentation Module

The right column is a batch image data enhancement with custom parameters (using the checked data increment method for all images in a folder with a certain probability)

• Parameter Initialization Module

Model basic parameters Select Configure

• Database Module

  

• Data Analysis Module

  

• Login Interface

  

YOUTUBE DEMO: Road Sign Recognition System Based on YOLOV5 BiliBili Demo: Road Sign Recognition System Based on YOLOV5

To install the required dependencies, run:

pip install -r requirements.txt

To run the application, you need to set up your MySQL database. Follow these steps to prepare your database:

• Automatic Database Creation (Optional):
  • If you prefer an automated setup, a batch script is provided. Run the setup_database.bat script to create the database. This requires MySQL to be installed and configured on your system.
• Manual Database Creation:
  • Alternatively, you can manually create the database in MySQL. Import and execute the data/regn_mysql.sql file in your MySQL environment to set up the necessary database and tables.

After setting up the database, update the connection Settings in the code to change the authentication information for your local database (these four variables are at the beginning of the code, around line 59, as follows); P.S. These authentication messages are called twice in the code (around lines 111 and 1783)

# Database connection settings as global variables
DB_HOST = 'localhost'    # Database host
DB_USER = 'root'         # Database user
DB_PASSWORD = '1234'     # Database password
DB_NAME = 'traffic_sign_recognition'  # Database name

If you encounter a RuntimeError: 'cryptography' package is required for sha256_password or caching_sha2_password auth methods, This is because the database authentication has gone wrong and the database needs to be properly created and the password entered. This error will also be reported if you do not have the mysql service started locally, so make sure your mysql service is started.

Here are the default login credentials:

Or modify the main function in main.py: remove the logon logic to enter the system directly without authentication.

• pt folder: Contains the YOLOv5 model file best.pt for road sign recognition.
• main_with folder: Contains login.py for the login UI and win.py for the main UI.
• dialog folder: Contains the RTSP pop-up interface.
• apprcc_rc.py: The resource file for the project.
• login_ji.py: Implements the login logic for the UI.
• data/run/run-exp52: The YOLOv5 road sign recognition model trained for 300 epochs.
• utils/tt100k_to_voc-main folder: Tool for converting JSON annotations to YOLO format.
• result: Folder to save inference results.
• run: Folder to save training logs and outputs.
• Dataset: Download from TT100k : Traffic-Sign Detection and Classification in the Wild.
• Database files: Located in the data folder, see -regn_mysql.sql for setup.

Since this project was done while I was learning YOLOv5 (quite a while ago), the main logic is concentrated in the main.py file. In other words, I didn't modularize different functions, and I didn't have a clear division of module structure. Now I want to divide it into modules, but I'm too lazy, ha ha 😄. If you're interested, you can modularize it so it's clearer.

• For converting the TT100K dataset to VOC format and selecting more than 100 images and XMLs for each category, see this CSDN blog post.
• The PyQt5-YOLOv5 integration was inspired by this GitHub repository.

WeChat：AIGCSD

email: [email protected]

Official Accounts：AI进修生

Track the GitHub star history of this project: