/PyroAId

Pyro AId is a computer vision project focused on detecting and analyzing fire-related events using YOLOv8. It predicts fires in videos, images, and live camera feeds.

Primary LanguageJupyter Notebook

Pyro AId

Pyro AId is a project aimed at detecting and analyzing fire-related events using YOLOv8, an advanced object detection model. It leverages computer vision to make predictions on video files, images, and live camera feeds.

Project Structure

Training

YOLOv8_training.ipynb:

  • Description: This Jupyter notebook trains a YOLOv8 model using a dataset from Roboflow.
  • Features:
    • Imports required libraries.
    • Initializes the dataset.
    • Trains the YOLOv8 model.
    • Visualizes training results.
    • Validates the trained model.

Prediction

YOLOv8_prediction.py:

  • Description: This Python script uses the trained YOLOv8 model to make predictions on video files, images, or live camera feeds.
  • Features:
    • Initializes the YOLO model with trained weights.
    • Allows for predictions on different sources (camera, video, image).
    • Configurable prediction parameters (e.g., confidence threshold, image size).

Requirements

  • ultralytics: For YOLOv8 utilities and model operations.
  • roboflow: For dataset management and downloading.

Install the necessary libraries using pip:

pip install ultralytics
pip install roboflow

Getting Started

  1. Training:

    • Replace TOKEN_PLACEHOLDER in YOLOv8_training.ipynb with your Roboflow API key.
    • Run the notebook to train the YOLOv8 model.

  2. Prediction:

    • Ensure the trained weights (best.pt) are available.
    • Update the paths for video, image, or camera in YOLOv8_prediction.py.

Running the Prediction Script

  1. Edit the YOLOv8_prediction.py file:

    • Set the video_path, image_path, or camera_path to the desired input source.

  2. Execute the script:

    python YOLOv8_prediction.py

Feel free to share and contribute!