A dataset of fortunella margarita images for object detection of deep learning based methods.

Abstract

Crops require appropriate planting techniques at different growth stages. Judgments on crop maturity affect the yield of crops. The planting and management of crops rely heavily on experienced farmers, which can reduce planting costs and increase yields. With the advancement of smart agriculture [1], images of crops can be used to accurately determine the growth stage of crops and estimate crop yields [2]. This can be combined with drones or smartphones to predict the growth stage and yield of Fortunella margarita for farmers in the future. This article presents an F. margarita image dataset. We classified F. margarita into three growth stages: mature, immature, and growing. In this dataset, an image may contain plants in several growth stages. The images were divided into seven categories according to growth stage. The dataset contains a total of 1031 original images. The total number of images was increased to 6611 through data augmentation. In addition, the dataset includes 6611 annotations with 7 categories of manually marked positions of F. margarita . Field images were captured in Jiaoxi, Yilan County, Taiwan, using smartphones. The dataset can serve as a resource for researchers who use different algorithms of machine learning or deep learning for object detection, image segmentation, and multiclass classification.

Plant phenotyping relevance

植物の生育段階を画像から判定するための注釈付きデータセットを提供しており、植物状態の画像取得・分類基盤が中心です。

Code and data availability

The paper's core asset is its own public dataset of 1031 original (6611 augmented) Fortunella margarita field images with 6611 XML annotations, deposited on Mendeley Data with explicit URLs. labelImg is a generic third-party tool, not a paper-specific asset.