Fire-Detection-Engine-ONNX

The Fire-Detection-Engine is a state-of-the-art deep learning model designed to detect fire-related conditions in images. It leverages the Vision Transformer (ViT) architecture, specifically the google/vit-base-patch16-224-in21k model, fine-tuned on a dataset of fire and non-fire images. The model is trained to classify images into one of the following categories: "Fire Needed Action," "Normal Conditions," or "Smoky Environment," making it a powerful tool for detecting fire hazards.

Classification report:

                    precision    recall  f1-score   support

Fire Needed Action     0.9708    0.9864    0.9785       808
 Normal Conditions     0.9872    0.9530    0.9698       808
 Smoky Environment     0.9818    1.0000    0.9908       808

          accuracy                         0.9798      2424
         macro avg     0.9799    0.9798    0.9797      2424
      weighted avg     0.9799    0.9798    0.9797      2424

Mappers

Mapping of IDs to Labels: {0: 'Fire Needed Action', 1: 'Normal Conditions', 2: 'Smoky Environment'} 

Mapping of Labels to IDs: {'Fire Needed Action': 0, 'Normal Conditions': 1, 'Smoky Environment': 2}

Key Features

• Architecture: Vision Transformer (ViT) - google/vit-base-patch16-224-in21k.
• Input: RGB images resized to 224x224 pixels.
• Output: Binary classification ("Fire Needed Action" or "Normal Conditions" or "Smoky Environment").
• Training Dataset: A curated dataset of fire place conditions.
• Fine-Tuning: The model is fine-tuned using Hugging Face's Trainer API with advanced data augmentation techniques.
• Performance: Achieves high accuracy and F1 score on validation and test datasets.