jlext07/jaguars

---
dataset_info:
features:
- name: filename
dtype: string
- name: label
dtype: string
- name: image
dtype: image
splits:
- name: raw_images
num_bytes: 34945123504.804
num_examples: 3098
- name: cropped_body
num_bytes: 2060540969.592
num_examples: 3098
- name: cropped_head
num_bytes: 737914072.544
num_examples: 3098
- name: segmented_body
num_bytes: 1265192956.906
num_examples: 3098
download_size: 38280262700
dataset_size: 39008771503.846
configs:
- config_name: default
data_files:
- split: raw_images
path: data/raw_images-*
- split: cropped_body
path: data/cropped_body-*
- split: cropped_head
path: data/cropped_head-*
- split: segmented_body
path: data/segmented_body-*
task_categories:
- image-classification
- image-segmentation
size_categories:
- 1K<n<10K
license: cc-by-4.0
pretty_name: jaguar_identification
language:
- en
tags:
- biology
---

# Jaguar Re-identification Dataset

This dataset contains images of jaguars from the Porto Jofre region in the Pantanal National Park, Brazil. It was curated for the purpose of developing and evaluating deep learning models for individual jaguar identification for population tracking.


## Dataset Description

The [Jaguar Identification Project](https://www.jaguaridproject.com/) aims to track jaguar movements, health, and demographics. This contributes valuable data to conservation strategies, especially considering jaguars are classified as Near Threatened by the [IUCN](https://iucn.org/) due to habitat loss, poaching, and human-wildlife conflict.

The core idea is to use photos taken by citizen scientists, to distinguish individual jaguars (e.g., identifying "Medrosa" vs. "Patricia").

**Origin of Data:**
The images are primarily sourced from Porto Jofre, a wildlife reserve located in the Pantanal National Park, which is home to one of the largest wild jaguar populations and is a popular destination for ecotourism and wildlife observation.

**Goal of the Dataset:**
The primary goal is to facilitate the development of models that can distinguish different individual jaguars from images. This supports:
* Population tracking and demographic studies.
* Citizen science efforts
* Conservation strategies by providing data on jaguar movements and health.

## Dataset Details

**Data Collection and Preprocessing:**
The dataset was prepared through a pipeline involving:
* Collection of raw images (initially around 40GB, ~4300 images).
* **Cropping and Segmentation:**
* Cropped body images (using Grounding DINO).
* Cropped face images.
* Segmented body images (using SAM - Segment Anything Model).
* Segmented face images.

The dataset includes images of various individual jaguars, such as "Ousado," "Medrosa," and "Patricia."

## Motivation

Protecting endangered and near-threatened species like jaguars starts with understanding their populations and the threats they face, such as wildfires, deforestation, illegal trade, climate change, and human-wildlife conflicts.

## Project Goals and Model Development

The dataset is used to:
* Contribute to citizen science efforts at the Pantanal Jaguar ID Project.
* Support research on re-identification models.

## Citation and License

All images were captured by Abigail Martin and contributors to the Jaguar ID Project https://www.jaguaridproject.com. The rights of ownership to these images remain to the Jaguar ID Project.
We kindly ask you to keep all credits as shown on the original images. You will find these in the `raw_images` split.

If you use this dataset in your research, please cite the following:

```bibtex
@misc{jaguar_identification_dataset_2025,
author = {Antonio Rueda-Toicen and Abigail Martin and Shahabeddin Dayani and Davide Panza and Aleksandra Kudaeva and Gerard de Melo},
title = {Jaguars of Pantanal Re-Identification Dataset},
year = {2025},
publisher = {Hugging Face},
journal = {Hugging Face Hub},
howpublished = {\url{https://huggingface.co/datasets/jaguaridentification/jaguars}}
}

```

数据集信息：
特征：
- 字段名：filename，数据类型：字符串（string）
- 字段名：label，数据类型：字符串（string）
- 字段名：image，数据类型：图像（image）
划分集：
- 划分名：raw_images（原始图像），字节数：34945123504.804，样本数：3098
- 划分名：cropped_body（裁剪躯体图像），字节数：2060540969.592，样本数：3098
- 划分名：cropped_head（裁剪头部图像），字节数：737914072.544，样本数：3098
- 划分名：segmented_body（分割躯体图像），字节数：1265192956.906，样本数：3098
下载大小：38280262700字节，数据集总大小：39008771503.846字节
配置：
- 配置名：default（默认配置），数据文件：
- 划分：raw_images（原始图像），路径：data/raw_images-*
- 划分：cropped_body（裁剪躯体图像），路径：data/cropped_body-*
- 划分：cropped_head（裁剪头部图像），路径：data/cropped_head-*
- 划分：segmented_body（分割躯体图像），路径：data/segmented_body-*
任务类别：图像分类（image-classification）、图像分割（image-segmentation）
样本规模类别：1K < n < 10K
许可证：cc-by-4.0
美观名称：jaguar_identification（美洲豹识别数据集）
语言：英语（en）
标签：生物学（biology）

# 美洲豹重识别数据集
本数据集收录了巴西潘塔纳尔国家公园波尔图若弗雷区域的美洲豹影像，专为开发与评估用于个体美洲豹识别以开展种群追踪的深度学习模型而构建。

## 数据集说明
[美洲豹识别项目](https://www.jaguaridproject.com/)旨在追踪美洲豹的活动、健康状况与种群结构，为保护策略提供宝贵数据支撑。鉴于美洲豹因栖息地丧失、偷猎及人兽冲突被国际自然保护联盟（IUCN，https://iucn.org/）列为近危物种，本数据集的价值尤为突出。
本数据集的核心思路是利用公民科学家拍摄的影像，实现美洲豹个体的区分（例如识别"Medrosa"与"Patricia"两只个体）。

**数据来源**：
影像主要采集自潘塔纳尔国家公园内的野生动物保护区波尔图若弗雷区域，该区域拥有全球规模最大的野生美洲豹种群之一，同时也是生态旅游与野生动物观测的热门目的地。

**数据集目标**：
本数据集的核心目标是助力能够从影像中区分不同美洲豹个体的模型开发，具体支撑以下方向：
* 种群追踪与种群结构研究
* 公民科学项目
* 通过提供美洲豹活动与健康数据，支撑保护策略制定。

## 数据集详情
**数据采集与预处理**：
本数据集通过以下流程构建完成：
* 原始影像采集（初始规模约40GB，包含约4300张影像）
* **裁剪与分割**：
* 基于Grounding DINO生成的裁剪躯体影像
* 裁剪头部影像
* 基于SAM（Segment Anything Model）生成的分割躯体影像
* 分割头部影像
本数据集收录了多只美洲豹个体的影像，例如"Ousado""Medrosa"与"Patricia"。

## 项目动因
保护美洲豹这类濒危或近危物种的前提，是了解其种群现状与面临的威胁，包括野火、森林砍伐、非法贸易、气候变化及人兽冲突等。

## 项目目标与模型开发
本数据集可用于：
* 助力潘塔纳尔美洲豹识别项目的公民科学工作
* 支持重识别模型相关研究。

## 引用与许可
所有影像均由阿比盖尔·马丁（Abigail Martin）及美洲豹识别项目的贡献者拍摄，影像版权归美洲豹识别项目所有。
我们恳请您保留原始影像上标注的所有署名信息，该信息可在`raw_images`划分的文件中获取。

若您在研究中使用本数据集，请引用如下文献：
bibtex
@misc{jaguar_identification_dataset_2025,
author = {Antonio Rueda-Toicen and Abigail Martin and Shahabeddin Dayani and Davide Panza and Aleksandra Kudaeva and Gerard de Melo},
title = {Jaguars of Pantanal Re-Identification Dataset},
year = {2025},
publisher = {Hugging Face},
journal = {Hugging Face Hub},
howpublished = {url{https://huggingface.co/datasets/jaguaridentification/jaguars}}
}