This project utilizes YOLOv11, a state-of-the-art object detection model, to identify fire in various contexts, including images, videos, and live webcam feeds. YOLOv11 offers advancements in detection accuracy, speed, and scalability, making it ideal for real-time applications in safety monitoring and disaster prevention.
The model is trained using a curated dataset of fire images, with annotations tailored for object detection. Tools like Roboflow were used for preprocessing and augmentation.
The base model, yolo11n.pt, is fine-tuned on the fire dataset. The configuration was optimized to balance performance and computational efficiency.
The model was trained with:
- Hyperparameters: Tuned for optimal performance on the fire dataset.
- Loss Functions: To ensure accurate localization and classification.
- Validation: A separate dataset was used to evaluate performance and adjust parameters.
The trained YOLOv11 model demonstrates robust fire detection on:
- Static Images: Quick identification of fire in photos.
- Video Feeds: Continuous detection in surveillance footage.
- Live Webcam Feeds: Real-time fire detection for dynamic scenarios.
The YOLOv11 fire detection system achieves high precision and recall, ensuring reliable identification across different environments.
- Fire monitoring in forests and remote areas.
- Industrial safety inspections.
- Enhancing smart surveillance systems.