Determining flower colors from images using artificial intelligence

Ron Wehrens; Manya Afonso; Hubert Fonteijn; Joao Paulo; Gerrit Polder; Marcel Rijsbergen; Gerard van Hameren; Raoul Haegens; Mary van den Helder; Henk Zwinkels

doi:10.1007/s10681-023-03258-2

Determining flower colors from images using artificial intelligence

Ron Wehrens^*, Manya Afonso, Hubert Fonteijn, Joao Paulo, Gerrit Polder, Marcel Rijsbergen, Gerard van Hameren, Raoul Haegens, Mary van den Helder, Henk Zwinkels

^*Corresponding author for this work

Research output: Contribution to journal › Article › Academic › peer-review

1 Citation (Scopus)

Abstract

Flower color is one of the most important traits in horticulture, and is one of the characteristics recorded to describe new varieties. In this paper, we examine four large real-world databases of roses and gerberas containing both images and color descriptions, and use state-of-the-art methods to automatically extract color descriptions from the images. Both Deep Learning and methods based on color histograms lead to success rates of approximately 85%. Deep learning has the advantage that no preprocessing is necessary—the more traditional methods lead to additional insight in the final color classification.

Original language	English
Article number	6
Number of pages	13
Journal	Euphytica
Volume	220
Early online date	14 Dec 2023
DOIs	https://doi.org/10.1007/s10681-023-03258-2
Publication status	Published - Jan 2024

Keywords

Color histograms
Deep learning
Ornamentals
Plant variety testing
Random forests

Access to Document

10.1007/s10681-023-03258-2

https://edepot.wur.nl/646533Licence: Taverne

Cite this

@article{40a70142134f44f98ada55c725c53c6d,

title = "Determining flower colors from images using artificial intelligence",

abstract = "Flower color is one of the most important traits in horticulture, and is one of the characteristics recorded to describe new varieties. In this paper, we examine four large real-world databases of roses and gerberas containing both images and color descriptions, and use state-of-the-art methods to automatically extract color descriptions from the images. Both Deep Learning and methods based on color histograms lead to success rates of approximately 85%. Deep learning has the advantage that no preprocessing is necessary—the more traditional methods lead to additional insight in the final color classification.",

keywords = "Color histograms, Deep learning, Ornamentals, Plant variety testing, Random forests",

author = "Ron Wehrens and Manya Afonso and Hubert Fonteijn and Joao Paulo and Gerrit Polder and Marcel Rijsbergen and {van Hameren}, Gerard and Raoul Haegens and {van den Helder}, Mary and Henk Zwinkels",

year = "2024",

month = jan,

doi = "10.1007/s10681-023-03258-2",

language = "English",

volume = "220",

journal = "Euphytica",

issn = "0014-2336",

publisher = "Springer",

}

TY - JOUR

T1 - Determining flower colors from images using artificial intelligence

AU - Wehrens, Ron

AU - Afonso, Manya

AU - Fonteijn, Hubert

AU - Paulo, Joao

AU - Polder, Gerrit

AU - Rijsbergen, Marcel

AU - van Hameren, Gerard

AU - Haegens, Raoul

AU - van den Helder, Mary

AU - Zwinkels, Henk

PY - 2024/1

Y1 - 2024/1

N2 - Flower color is one of the most important traits in horticulture, and is one of the characteristics recorded to describe new varieties. In this paper, we examine four large real-world databases of roses and gerberas containing both images and color descriptions, and use state-of-the-art methods to automatically extract color descriptions from the images. Both Deep Learning and methods based on color histograms lead to success rates of approximately 85%. Deep learning has the advantage that no preprocessing is necessary—the more traditional methods lead to additional insight in the final color classification.

AB - Flower color is one of the most important traits in horticulture, and is one of the characteristics recorded to describe new varieties. In this paper, we examine four large real-world databases of roses and gerberas containing both images and color descriptions, and use state-of-the-art methods to automatically extract color descriptions from the images. Both Deep Learning and methods based on color histograms lead to success rates of approximately 85%. Deep learning has the advantage that no preprocessing is necessary—the more traditional methods lead to additional insight in the final color classification.

KW - Color histograms

KW - Deep learning

KW - Ornamentals

KW - Plant variety testing

KW - Random forests

U2 - 10.1007/s10681-023-03258-2

DO - 10.1007/s10681-023-03258-2

M3 - Article

AN - SCOPUS:85180170063

SN - 0014-2336

VL - 220

JO - Euphytica

JF - Euphytica

M1 - 6

ER -

Determining flower colors from images using artificial intelligence

Abstract

Keywords

Access to Document

Fingerprint

Cite this