Data from the paper DeepPhenoTree - Apple Edition

DeepPhenoTree - Apple Edition Images of apple trees taken in four European sites (Switzerland, Belgium, Spain, Italy) in 2025. Images are grouped in three sets associated to phenological scales : flowering, fruitlet and fruit. These images are supplementary material of the paper "DeepPhenoTree - Apple Edition : a Multi-site apple phenology RGB annotated dataset with deep learning baseline models". Overview The data corresponds to the dataset associated with the best model designed in the paper: Metuarea Herearii, Ousseini Hamza Abdoul-Djalil, Guerra Walter, Zuffa Francesca, Panzeri Francesco, Patocchi Andrea, Lozano Lidia, Van Hoye Shauny, Laurens François, Labrosse Jeremy, Rasti Pejman & Rousseau David. (2026, February 27). DeepPhenoTree – Apple Edition: A multi-site apple phenology RGB annotated dataset with deep learning baseline models (Version 1) [Preprint]. Research Square. https://doi.org/10.21203/rs.3.rs-8977752/v1 The data originates from three source datasets: Flowering Fruitlet Fruit All images and labels have been anonymized. Each sample is associated with: an anonymous file name a source country among Belgium, Switzerland, Italy, Spain a final split in the manifest: train, val Directory Structure DeepPhenoTree - Apple Edition/ ├── Flowering.zip/ │ ├── images/ │ │ ├── train/ │ │ └── val/ │ └── labels/ │ ├── train/ │ └── val/ ├── Fruitlet.zip/ │ ├── images/ │ │ ├── train/ │ │ └── val/ │ └── labels/ │ ├── train/ │ └── val/ ├── Fruit.zip/ │ ├── images/ │ │ ├── train/ │ │ └── val/ │ └── labels/ │ ├── train/ │ └── val/ ├── Fruit.csv ├── Fruitlet.csv ├── Flowering.csv └── README.md Compressed files contain the data (images and labels). Table files (.csv) contain the locations where the images were taken. They are structured as follows: image location split The image column corresponds to the anonymized image name. README.md provides a description of the repository DeepPhenoTree - Apple Edition. Label Convention Labels are stored in YOLO text format. The first column is rewritten so that the class ID depends on the dataset: 0 for Flowering 1 for Fruitlet 2 for Fruit The remaining bounding-box coordinates are preserved. Citing DeepPhenoTree - Apple Edition If you use the DeepPhenoTree - Apple Edition dataset in your research, please use the following BibTeX entry. @dataset{metuarea2026deepphenotree_dataset, author = {Metuarea, Herearii and Abdoul-Djalil, Ousseini Hamza and Decastro, Lou and Marhadour, Jade and Karia, Oumaima and Masson, Lorène and Kourkoumelis-Rodostamos, Marie and Guerra, Walter and Zuffa, Francesca and Panzeri, Francesco and Patocchi, Andrea and Lozano, Lidia and Van Hoye, Shauny and Rymenants, Marijn and Laurens, François and Labrosse, Jeremy and Rasti, Pejman and Rousseau, David}, title = {DeepPhenoTree – Apple Edition}, year = {2026}, version = {5}, publisher = {Recherche Data Gouv}, doi = {10.57745/NORPF1} }