---
library_name: keras
pipeline_tag: image-classification
tags:
  - keras
  - tensorflow
  - computer-vision
  - image-classification
  - fire-detection
  - fire-severity
  - severity-classification
  - multi-class-classification
  - transfer-learning
---

# Fire Severity Classification Models

A collection of trained deep learning models for three-class fire severity classification.

The models classify fire images into three severity categories:

- **Mild**
- **Moderate**
- **Severe**

Multiple CNN architectures are included for model comparison.

## Models

| Model | Architecture | Task |
|---|---|---|
| `efficientnetb0_tri_classification.keras` | EfficientNetB0 | Mild / Moderate / Severe |
| `resnet50_tri_classification.keras` | ResNet50 | Mild / Moderate / Severe |
| `xception_tri_classification_finalized.keras` | Xception | Mild / Moderate / Severe |

## Model Format

All models are saved using the Keras `.keras` format.

```python
import tensorflow as tf

model = tf.keras.models.load_model(
    "efficientnetb0_tri_classification.keras"
)

model.summary()
```

Replace the filename with the model you want to load.

## Classification Task

```text
Input Image
     |
     v
CNN Model
     |
     v
Fire Severity
     |
     +---- Mild
     |
     +---- Moderate
     |
     +---- Severe
```

## Intended Use

These models are intended for:

- research
- experimentation
- educational purposes
- model comparison
- further fine-tuning
- fire severity analysis

## Limitations

Model performance depends on the training dataset, preprocessing pipeline, image quality, class distribution, and deployment environment.

These models should be independently evaluated before being used in real-world or safety-critical applications.

## Project Context

These models are part of a larger deep learning project involving fire detection, fire severity classification, feature extraction, dimensionality reduction, clustering, and recommendation generation.

## License

No standardized open-source license has been specified for this repository.

Please refer to the original project and dataset terms before redistribution or commercial use.