---
license: apache-2.0
datasets:
- odiug77/wildfire
base_model:
- Ultralytics/YOLO26
pipeline_tag: object-detection
tags:
- wildfire
- smoke
- fire
- yolo
- bounding-box
- yolo26
---

# Wildfire Smoke and Flame Detection – YOLOv26m (Updated)

## Overview

This model is based on the **YOLOv26m** architecture and has been specifically fine-tuned to detect **wildfire smoke and flames** in outdoor environments.

Compared to previous versions, this update utilizes a larger dataset and optimized training hyperparameters, resulting in a significant boost in **Recall** and **mAP**. It is designed for early wildfire detection using standard RGB surveillance cameras positioned at strategic observation points such as hills, fire towers, or remote monitoring stations.

## 🎯 Purpose

The primary goal of this model is **early wildfire detection**, enabling faster response times and reducing environmental and economic damage. 

Optimized for:
- **Daytime & Nighttime** wildfire detection.
- **Smoke Recognition:** Critical for long-distance early detection.
- **Flame Detection:** For immediate confirmation of fire events.
- **Standard RGB Imagery:** Works with standard optical cameras (no thermal sensors required).

## 🧠 Model Architecture

- **Architecture:** YOLOv26m (fused)
- **Task:** Object Detection
- **Classes:** `smoke`, `fire`
- **Layers:** 132
- **Parameters:** 20,350,994
- **GFLOPs:** 67.9
- **Input Resolution:** 640x640 (RGB)

## 📊 Model Performance (Latest Validation)

The model was trained for **80 epochs** using **Automatic Mixed Precision (AMP)** and **AdamW** optimizer. 

### Overall Metrics (Validation Set)
| Metric | Value | Improvement vs Previous |
| :--- | :--- | :--- |
| **Precision (P)** | 0.635 | +0.3% |
| **Recall (R)** | **0.681** | **+17.2%** |
| **mAP @0.50** | **0.675** | **+13.6%** |
| **mAP @0.50:0.95** | **0.425** | **+11.9%** |

### Per-Class Metrics
| Class | Images | Instances | Precision | Recall | mAP@0.50 | mAP@0.50:0.95 |
| :--- | :--- | :--- | :--- | :--- | :--- | :--- |
| **Smoke** | 180 | 253 | 0.633 | 0.767 | 0.752 | 0.491 |
| **Fire** | 123 | 836 | 0.637 | 0.596 | 0.598 | 0.359 |

### Inference Speed (Tesla T4 GPU)
- **Pre-process:** 0.3 ms
- **Inference:** 12.3 ms
- **Post-process:** 0.8 ms
- **Total Throughput:** ~74 FPS (theoretical)

## 🚀 Inference Example (CPU)

Minimal example to run inference on **CPU** and print detection results.

```python
from ultralytics import YOLO

# Load the trained model
model = YOLO("wildfire-smoke-fire.pt")

# Run inference on an image
results = model("forest_view.jpg", conf=0.25)

# Process results
for r in results:
    print(f"Detected {len(r.boxes)} objects.")
    r.show()  # Display annotated image
```


## 🚀 Usage Example

Install the required library:
```bash
pip install ultralytics
```

# 🛠️ Training Configuration
- Optimizer: AdamW (lr=0.001667, momentum=0.9)
- Augmentations: Blur, MedianBlur, ToGray, CLAHE
- Environment: PyTorch 2.1.0+cu128 on Tesla T4