/YOLOdetectify

Yolodetectify harnesses the power of the YOLOv8m model to enable efficient object detection on custom datasets.

Primary LanguagePythonMIT LicenseMIT

Yolodetectify

Description

Yolodetectify harnesses the power of the YOLOv8m model to enable efficient object detection on custom datasets. By seamlessly integrating the bing_photo_fetcher, it automates the process of sourcing and organizing internet images to fit the specific requirements of YOLOv8m. Furthermore, the tool is engineered to autonomously partition datasets into training and validation sets, and the yolo_config_manager ensures effortless creation of the essential data.yaml configuration for the training pipeline. Users can interact with the model either through a user-friendly CLI or directly as a script via model.py.

Features

  • Image Fetching: Utilize bing_photo_fetcher to source and format images directly from the web, optimized for YOLOv8m.
  • Data Partitioning: Automated splitting of datasets into training and validation sets ensures efficient model training.
  • YAML Configuration: With yolo_config_manager, generate the data.yaml configuration vital for the training pipeline.
  • Dual Invocation: Opt for either Command Line or script-based execution with model.py.
  • Dependencies Management: A comprehensive requirements.txt file streamlines environment setup.

Setup & Installation

  1. Begin by cloning this repository to your local workspace. 
    git clone gh repo clone etemkocaaslan/YOLOdetectify
  2. Move to the newly cloned project directory: 
    cd Yolodetectify
  3. Set up your environment by installing the necessary packages: 
    pip install -r requirements.txt

Usage

  • Bing Photo Fetcher: Download images based on search queries:

    python bing_photo_fetcher.py --search_queries Football soccer --num_images_per_query 100 --output_folder_name images

  • YOLO Config Manager: Generate YOLO configurations:

    python yolo_config_manager.py --train_path /path/to/train/data --val_path /path/to/val/data --nc 2 --names Player Ball

  • Model Training: To train the YOLO model:

    python model.py --model_name yolov8m.pt --data data.yaml --epochs 30 --imgsz 640

Note: Modify the paths and values in the above CLI commands according to your specific configurations.

Contributors

  • Etem Kocaaslan

License

Yolodetectify operates under MIT License. Detailed licensing information is available in the LICENSE file.