YOLO Object Detection Project

This project implements a YOLOv8 object detection model for identifying objects in video and image data.

Project Structure

The project consists of the following files and directories:

• .gitignore
• README.md
• requirements.txt: Lists the project dependencies.
• YOLO_inference.py: Contains the code for running inference on video files using the trained YOLOv8 model.
• YOLO_lossVisualization.py
• YOLO_train.py: Contains the code for training and evaluating the YOLOv8 model.
• data/: Contains the data-related files.
  • data_preprocessing.py: Contains the code for preprocessing video and image data.
• model/: Contains the model files.
  • 4class.ptthe model that can distinguish - closed box, open box, foods and hands
  • 11class.ptthe model that can distinguish - normal box, red box, black box and white box. Meanwhile those boxes can be distinguished by open/closed. Also, the food can being distinguished by normal/spicy Chiken and Tokbbokki
  • 9class.ptcompared to the 11class one, this one can only distinguish the food to food
  • yolov8nOG.ptoriginal yolov8n model

Dependencies

The project requires the following dependencies:

• ultralytics
• cv2
• lap
• pandas
• matplotlib

You can install these dependencies using pip:

pip install -r requirements.txt

Usage

1. Data Preparation

1. Place your raw video data in the data/raw_data/ directory.

2. Run the data/data_preprocessing.py script to preprocess the video data into image frames.

   python data/data_preprocessing.py

   This script will:

   • Extract frames from the videos in data/raw_data/.
   • Enhance the images for better object detection (optional).
   • Save the processed images in data/image_data/<video_name>/.

3. create alabelled_data4OBJ folder under the data directory. Label the images in YOLO format, and put them in labelled_data4OBJ/. Which should contain the following structure:

labelled_data4OBJ/
  ├── images/
  ├── labels/
  ├── classes.txt
  └── notes.json

2. Training

1. Modify the YOLO_train.py script to configure the training parameters, such as the number of epochs, batch size, and learning rate.

2. Run the YOLO_train.py script to train the YOLOv8 model.

   python YOLO_train.py

   This script will:

   • Split the dataset into training and testing sets.
   • Create a data.yaml file that defines the dataset configuration for YOLO.
   • Train the YOLOv8n model using the specified parameters.
   • Evaluate the trained model on the test set.
   • Save the trained model weights and training results in the model/ directory.

3. Inference

1. Modify the YOLO_inference.py script to specify the path to the video file you want to run inference on.

2. Run the YOLO_inference.py script to perform object detection on the video.

   python YOLO_inference.py

   This script will:

   • Load the trained YOLOv8 model weights from model/.
   • Run inference on the specified video file.
   • Display the results with bounding boxes around the detected objects.

---

Created by Yonsei Data Science Lab.