wildintelproject/donadataset

Doñana National Park Dataset (DNP_DS)

Version: 1.0

Camera-trap images from Doñana National Park (Spain) annotated with bounding boxes for object detection in YOLO format. It contains 33,914 images split into train, val and test, with 38,492 annotated objects of 15 animal categories (plus empty images): mammals identified to species or genus (e.g. Cervus elaphus, Dama dama, Sus scrofa, Vulpes vulpes, Lynx pardinus, Meles meles, Oryctolagus cuniculus), domestic animals (Bos taurus, Canis familiaris, Felis catus, Equus sp) and a general bird category (Ave). It is meant for training and evaluating wildlife detection models.

Dataset format

This dataset is distributed in YOLO training format: image files split into train/val/test sets, described by a data.yaml config.

It has 18 classes: Ave, Felis catus, Bos taurus, Canis familiaris, Herpestes ichneumon, Meles meles, Oryctolagus cuniculus, Dama dama, Genetta genetta, Equus sp, Homo sapiens, Lepus granatensis, Lynx pardinus, Cervus elaphus, Vulpes vulpes, Sus scrofa, Vehicle, Empty.

.
├── data.yaml             ← YOLO dataset config (splits, classes)
├── images/
│   ├── train/            ← training images
│   ├── val/               ← validation images
│   └── test/              ← test images (if present)
├── labels/                ← training labels (if present in the source dataset)
│   ├── train/
│   ├── val/
│   └── test/
├── checksums-sha256.txt
├── CITATION.cff
└── LICENSE

The train split is too large for a single Hugging Face Hub folder (10,000 files at most), so its images and labels are spread over hash-named subfolders (e.g. images/train/3f/img.jpg with its label at labels/train/3f/img.txt). YOLO trainers find them just the same: they look for images recursively under each split's folder.

Dataset statistics

Split	Images	Labeled images	Objects
train	27467	27467	31962
val	3264	3264	3133
test	3183	3183	3397

Class	Objects
Ave	2559
Felis catus	8
Bos taurus	6001
Canis familiaris	373
Herpestes ichneumon	844
Meles meles	2480
Oryctolagus cuniculus	1189
Dama dama	3862
Genetta genetta	398
Equus sp	7352
Homo sapiens	0
Lepus granatensis	1689
Lynx pardinus	28
Cervus elaphus	4521
Vulpes vulpes	3068
Sus scrofa	4120
Vehicle	0
Empty	0

Integrity verification

checksums-sha256.txt lists the SHA-256 hash of every file in this export, to verify nothing was corrupted or truncated after downloading.

Source

Generated via wildintel-publisher.

Funding

This work is part of the WildINTEL project, funded by the Biodiversa+ Joint Research Call 2022-2023 "Improved transnational monitoring of biodiversity and ecosystem change for science and society (BiodivMon)". Biodiversa+ is the European co-funded biodiversity partnership supporting excellent research on biodiversity with an impact for policy and society. Biodiversa+ is part of the European Biodiversity Strategy for 2030 that aims to put Europe's biodiversity on a path to recovery by 2030 and is co-funded by the European Commission.

WildINTEL has been co-funded by the European Commission (GA No. 101052342) and the following funding organisations: Agencia Estatal de Investigación (Spain, PCI2023-145963-2, PCI2024-153489), National Science Centre (Poland, UMO-2023/05/Y/NZ8/00104), the Research Council of Norway (Norway, NFR350962) and the German Research Foundation (Germany).

Additional support was provided by Fundación Biodiversidad, Ministerio para la Transición Ecológica y el Reto Demográfico, Project AI-CENSUS and the Programa Operativo FEDER Andalucía 2014–2020, Project UHU-202028.

Acknowledgements

We thank everyone who took part in the fieldwork, image collection and annotation behind this dataset, and the WildINTEL project partners for their support.

The authors acknowledge the University of Huelva for providing the computational infrastructure and technical resources used in this study. We also acknowledge the contributions of the WildINTEL consortium and the Biodiversa+ partnership, which provided the collaborative framework within which this research was conducted. The authors gratefully acknowledge Sara Beery and the contributors to the Caltech Camera Traps (CCT) dataset for making this valuable resource publicly available. Images from the Caltech Camera Traps dataset were used as part of the training data for model development; these images are not redistributed as part of the dataset accompanying this study.

License

This dataset is released under: CC-BY-NC-4.0. See the LICENSE file for the full text.

Citation

To cite this dataset:

Javier Calzada Samperio, Santiago Gutiérrez-Zapata, Simone Santoro, Manuel E. Gegundez-Arias, Iñaki Fernández de Viana, Lenka Straková, & Agata Kaczówka (2026). Doñana National Park Dataset (DNP_DS) (Version 1.0) [Data set]. University of Huelva. https://zenodo.org/record/23917216 © 2026 University of Huelva

This citation was generated from the CITATION.cff file. See About CITATION files for more info.

Limitations and ethical considerations

Before using this dataset, users should review possible limitations related to class imbalance, annotation quality, environmental bias, and any privacy or sensitive information the images may contain.