Related Experiment Video
Updated: Jul 5, 2026
11:02
Fruit Volatile Analysis Using an Electronic Nose
Published on: March 30, 2012
22.6K
Dataset for orange fruit detection from UAV in citrus orchards.
Guillem Montalban-Faet1, Enrique Navarro-Modesto1, Andoni Salcedo-Navarro1
1Computer Science Department, ETSE-UV, Universitat de València, València, Spain.
Data in Brief
|April 20, 2026
Summary
Researchers developed CampanetaOrangeFruit, a new dataset for orange fruit detection using multispectral UAV imagery. This resource aids in advancing precision agriculture and automated monitoring in citrus orchards.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Remote Sensing
Background:
- Accurate fruit detection is crucial for citrus orchard management, including yield estimation and precision harvesting.
- Existing datasets for orange fruit detection are limited, especially those using multispectral data under field conditions, hindering deep learning model development.
Purpose of the Study:
- To introduce CampanetaOrangeFruit, a novel, publicly available dataset for orange fruit detection.
- To provide synchronized RGB and multispectral (R, G, RE, NIR) UAV imagery with detailed annotations for benchmarking deep learning models.
Main Methods:
- Acquisition of data using a DJI Mavic 3 Multispectral UAV over a commercial citrus orchard.
- Generation of 2750 images from 550 synchronized captures, featuring over 300,000 annotated orange instances.
- Annotation in YOLOv5 format using a homography-based reprojection process for cross-spectral geometric consistency.
Main Results:
- The dataset contains pixel-aligned, cross-spectral UAV imagery suitable for various agricultural applications.
- CampanetaOrangeFruit enables research into cross-spectral and illumination-invariant fruit detection.
Conclusions:
- CampanetaOrangeFruit serves as a valuable benchmark for deep learning in precision agriculture.
- The dataset supports advancements in automated orchard monitoring and sustainable citrus production.
