High-Resolution UAV Datasets for Agricultural Computer Vision

Published: 24 April 2026| Version 2 | DOI: 10.17632/r4w3b2mfnw.2
Contributors:
, Mohamed Adnane Mahraz, Ali Yahyaouy, Ali Achebour, Jamal Riffi, Hamid Tairi

Description

A high-resolution UAV dataset for computer vision applications in precision agriculture. It includes multi-crop imagery from olive, apricot, and vineyard fields, along with annotations and orthomosaic products. The dataset supports tasks such as object detection, segmentation, classification, and GIS-based analysis.

Files

Steps to reproduce

1. Download the TreeCrownsDb dataset and extract all files. 2. Organize the dataset into images, annotations, orthomosaic, and metadata folders. 3. Extract metadata (date, time, GPS coordinates, altitude, and acquisition conditions) from the image files. 4. Generate orthomosaics using photogrammetric software (e.g., Pix4D or Agisoft Metashape) if not provided. 5. Load the dataset into a computer vision framework (e.g., PyTorch, TensorFlow, or YOLO). 6. Preprocess the data (resizing, normalization, and augmentation if needed). The images can be used directly or divided into tiles (patches) of appropriate size depending on the training requirements. 7. Train and evaluate models for object detection, segmentation, and classification tasks. 8. Optionally, use the orthomosaic data for GIS-based analysis and large-scale mapping.

Categories

Computer Vision, Image Acquisition, Image Segmentation, Object Detection, Image Classification, Orthophoto, Deep Learning, Image Analysis, Object Detection Algorithm

Licence