---
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
pretty_name: VanRakshak Forest & Wildlife Aerial-Thermal Dataset
language:
- en
license: apache-2.0
task_categories:
- object-detection
tags:
- object-detection
- computer-vision
- drone
- uav
- thermal-imaging
- wildfire
- wildlife-conservation
- anti-poaching
- forestry
- yolov8
size_categories:
- 1K<n<10K
---

# 🌲 VanRakshak: Forest & Wildlife Aerial-Thermal Dataset

[![GitHub Repository](https://img.shields.io/badge/GitHub-sanjeevafk%2Fvanrakshak-black?logo=github)](https://github.com/sanjeevafk/vanrakshak)
[![Trained Model](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Model%20(YOLOv8n)-yellow)](https://huggingface.co/sanjeevafk/vanrakshak-forest-yolov8n)
[![Colab Notebook](https://img.shields.io/badge/Google%20Colab-Train%20Model-orange?logo=googlecolab)](https://colab.research.google.com/github/sanjeevafk/vanrakshak/blob/main/vanrakshak_yolov8_training.ipynb)
[![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
[![Dataset Viewer](https://img.shields.io/badge/Hugging%20Face-Dataset%20Viewer%20Ready-brightgreen)](https://huggingface.co/datasets/sanjeevafk/vanrakshak-forest-aerial-thermal)

A comprehensive, curated dataset of aerial and thermal imagery optimized for forest surveillance, anti-poaching, human-wildlife conflict mitigation, and wildfire detection via UAVs and drones.

---

## 📊 Quick Start

```python
from datasets import load_dataset

# Load full dataset with instant streaming and native image/bbox decoding
dataset = load_dataset("sanjeevafk/vanrakshak-forest-aerial-thermal")

# Access sample with bounding boxes and classes
sample = dataset["train"][0]
image = sample["image"]
bboxes = sample["objects"]["bbox"]
categories = sample["objects"]["category"]
```

### Target Classes (6 Categories)
| Class ID | Label | Description & Strategic Application |
|---|---|---|
| `0` | **person** | Human detection in dense canopy / thermal night patrols (anti-poaching & lost hikers) |
| `1` | **vehicle** | Unauthorized surveillance / patrol trucks / civilian intrusions |
| `2` | **timber_truck** | Illegal logging and illicit timber transit tracking |
| `3` | **fire** | Early-stage wildfire and active flame detection |
| `4` | **smoke** | Forest canopy smoke plumes for rapid wildfire alerting |
| `5` | **elephant** | Wildlife monitoring and human-elephant conflict mitigation |

---

## 📈 Dataset Statistics

| Split | Images | Bounding Boxes | Formats Available |
|---|---|---|---|
| **Train** | 4,000 | ~32,000+ | Parquet |
| **Validation** | 3,946 | ~31,000+ | Parquet |
| **Total** | **7,946** | **~63,000+** | Native Hugging Face & YOLOv8 |

---

## 📜 Citation & License

This dataset is distributed under the **Apache 2.0 License**.

```bibtex
@dataset{vanrakshak_2026,
  author       = {Sanjeev and VanRakshak Project Contributors},
  title        = {VanRakshak: Forest & Wildlife Aerial-Thermal Dataset},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\\url{https://huggingface.co/datasets/sanjeevafk/vanrakshak-forest-aerial-thermal}}
}
```