Taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi-task semantic segmentation

Herbicide research and development necessitate specific trials to monitor the effects of various herbicide formulations, quantities, and protocols on different plant species and growth stages. These trials are necessary to ensure the safety and efficacy of the developed products. Currently, these te...

Full description

Saved in:
Bibliographic Details
Main Authors: Artzai Picon, Daniel Mugica, Itziar Eguskiza, Arantza Bereciartua-Perez, Javier Romero, Carlos Javier Jimenez, Christian Klukas, Laura Gomez-Zamanillo, Till Eggers, Ramon Navarra-Mestre
Format: Article
Language:English
Published: Elsevier 2025-03-01
Series:Smart Agricultural Technology
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2772375524003654
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850041576995160064
author Artzai Picon
Daniel Mugica
Itziar Eguskiza
Arantza Bereciartua-Perez
Javier Romero
Carlos Javier Jimenez
Christian Klukas
Laura Gomez-Zamanillo
Till Eggers
Ramon Navarra-Mestre
author_facet Artzai Picon
Daniel Mugica
Itziar Eguskiza
Arantza Bereciartua-Perez
Javier Romero
Carlos Javier Jimenez
Christian Klukas
Laura Gomez-Zamanillo
Till Eggers
Ramon Navarra-Mestre
author_sort Artzai Picon
collection DOAJ
description Herbicide research and development necessitate specific trials to monitor the effects of various herbicide formulations, quantities, and protocols on different plant species and growth stages. These trials are necessary to ensure the safety and efficacy of the developed products. Currently, these tests are conducted manually and assessed visually, making the process time-consuming and labor-intensive. Developing a computer model to characterize species, damage, and growth stages is challenging due to the fine-grained differences between species and damage, significant intra-class variability, and difficulties in manual annotations. Additionally, manually annotated datasets for semantic segmentation are often imperfect. The presence of non-target or unknown species, where only the genus or family is known, complicates the management and scalability of these datasets.In this work, we propose a new hierarchical loss function, suitable for semantic segmentation tasks, capable to take advantage for the hierarchical taxonomy relationships between species, plant damages and other relationships and thus, reduce the need for annotated data. The proposed loss function support datasets with varying granularity and annotation heterogeneity, including for partial annotations at the pixel level. We validated this loss function using a multi-task semantic segmentation neural network to simultaneously detect plant species and quantify the damage of each species. The proposed hierarchical loss function improves model performance, increasing the F1-Score for species detection from 0.41 to 0.52, for damage detection from 0.23 to 0.28. This enhancement forces the model to learn richer hierarchical representations, enabling the support of heterogeneous and partially annotated scalable datasets, which are common in real-world AI applications.
format Article
id doaj-art-36192e6411474c659eb0deaeb7f7b660
institution DOAJ
issn 2772-3755
language English
publishDate 2025-03-01
publisher Elsevier
record_format Article
series Smart Agricultural Technology
spelling doaj-art-36192e6411474c659eb0deaeb7f7b6602025-08-20T02:55:45ZengElsevierSmart Agricultural Technology2772-37552025-03-011010076110.1016/j.atech.2024.100761Taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi-task semantic segmentationArtzai Picon0Daniel Mugica1Itziar Eguskiza2Arantza Bereciartua-Perez3Javier Romero4Carlos Javier Jimenez5Christian Klukas6Laura Gomez-Zamanillo7Till Eggers8Ramon Navarra-Mestre9TECNALIA, Basque Research and Technology Alliance (BRTA), Parque Tecnológico de Bizkaia, C/ Geldo. Edificio 700, E-48160 Derio - Bizkaia, Spain; University of the Basque Country, Plaza Torres Quevedo, 48013 Bilbao, Spain; Corresponding author at: TECNALIA, Basque Research and Technology Alliance (BRTA), Parque Tecnológico de Bizkaia, C/ Geldo. Edificio 700, E-48160 Derio - Bizkaia, Spain.TECNALIA, Basque Research and Technology Alliance (BRTA), Parque Tecnológico de Bizkaia, C/ Geldo. Edificio 700, E-48160 Derio - Bizkaia, SpainTECNALIA, Basque Research and Technology Alliance (BRTA), Parque Tecnológico de Bizkaia, C/ Geldo. Edificio 700, E-48160 Derio - Bizkaia, SpainTECNALIA, Basque Research and Technology Alliance (BRTA), Parque Tecnológico de Bizkaia, C/ Geldo. Edificio 700, E-48160 Derio - Bizkaia, SpainBASF Espanola S.L. Carretera A376, 41710 Utrera, Sevilla, SpainBASF Espanola S.L. Carretera A376, 41710 Utrera, Sevilla, SpainBASF SE, Speyererstrasse 2, 67117 Limburgerhof, GermanyTECNALIA, Basque Research and Technology Alliance (BRTA), Parque Tecnológico de Bizkaia, C/ Geldo. Edificio 700, E-48160 Derio - Bizkaia, SpainBASF SE, Speyererstrasse 2, 67117 Limburgerhof, GermanyBASF SE, Speyererstrasse 2, 67117 Limburgerhof, GermanyHerbicide research and development necessitate specific trials to monitor the effects of various herbicide formulations, quantities, and protocols on different plant species and growth stages. These trials are necessary to ensure the safety and efficacy of the developed products. Currently, these tests are conducted manually and assessed visually, making the process time-consuming and labor-intensive. Developing a computer model to characterize species, damage, and growth stages is challenging due to the fine-grained differences between species and damage, significant intra-class variability, and difficulties in manual annotations. Additionally, manually annotated datasets for semantic segmentation are often imperfect. The presence of non-target or unknown species, where only the genus or family is known, complicates the management and scalability of these datasets.In this work, we propose a new hierarchical loss function, suitable for semantic segmentation tasks, capable to take advantage for the hierarchical taxonomy relationships between species, plant damages and other relationships and thus, reduce the need for annotated data. The proposed loss function support datasets with varying granularity and annotation heterogeneity, including for partial annotations at the pixel level. We validated this loss function using a multi-task semantic segmentation neural network to simultaneously detect plant species and quantify the damage of each species. The proposed hierarchical loss function improves model performance, increasing the F1-Score for species detection from 0.41 to 0.52, for damage detection from 0.23 to 0.28. This enhancement forces the model to learn richer hierarchical representations, enabling the support of heterogeneous and partially annotated scalable datasets, which are common in real-world AI applications.http://www.sciencedirect.com/science/article/pii/S2772375524003654Taxonomic hierarchical lossPlant species and damage segmentationPrecision agriculturePrecise phenotypingDeep learning
spellingShingle Artzai Picon
Daniel Mugica
Itziar Eguskiza
Arantza Bereciartua-Perez
Javier Romero
Carlos Javier Jimenez
Christian Klukas
Laura Gomez-Zamanillo
Till Eggers
Ramon Navarra-Mestre
Taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi-task semantic segmentation
Smart Agricultural Technology
Taxonomic hierarchical loss
Plant species and damage segmentation
Precision agriculture
Precise phenotyping
Deep learning
title Taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi-task semantic segmentation
title_full Taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi-task semantic segmentation
title_fullStr Taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi-task semantic segmentation
title_full_unstemmed Taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi-task semantic segmentation
title_short Taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi-task semantic segmentation
title_sort taxonomic hierarchical loss function for enhanced crop and weed phenotyping in multi task semantic segmentation
topic Taxonomic hierarchical loss
Plant species and damage segmentation
Precision agriculture
Precise phenotyping
Deep learning
url http://www.sciencedirect.com/science/article/pii/S2772375524003654
work_keys_str_mv AT artzaipicon taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT danielmugica taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT itziareguskiza taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT arantzabereciartuaperez taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT javierromero taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT carlosjavierjimenez taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT christianklukas taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT lauragomezzamanillo taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT tilleggers taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation
AT ramonnavarramestre taxonomichierarchicallossfunctionforenhancedcropandweedphenotypinginmultitasksemanticsegmentation