Automated generation of ground truth images of greenhouse-grown
plant shoots using a GAN approach

J 2025

Automated generation of ground truth images of greenhouse-grown plant shoots using a GAN approach

ULLAH, Sajid; Narendra NARISETTI; Kerstin NEUMANN; Thomas ALTMANN; Jan HEJÁTKO et al.

Základní údaje

Originální název

Automated generation of ground truth images of greenhouse-grown plant shoots using a GAN approach

Autoři

ULLAH, Sajid; Narendra NARISETTI; Kerstin NEUMANN; Thomas ALTMANN; Jan HEJÁTKO a Evgeny GLADILIN

Vydání

PLANT METHODS, LONDON, BMC, 2025, 1746-4811

Další údaje

Jazyk

angličtina

Typ výsledku

Článek v odborném periodiku

Obor

10611 Plant sciences, botany

Stát vydavatele

Velká Británie a Severní Irsko

Utajení

není předmětem státního či obchodního tajemství

Odkazy

URL

Impakt faktor

Impact factor: 4.400 v roce 2024

Označené pro přenos do RIV

Ano

Kód RIV

RIV/00216224:14740/25:00143307

Organizační jednotka

Středoevropský technologický institut

Klíčová slova anglicky

High-throughput greenhouse imaging; Plant phenotyping; Image segmentation; Ground truth data generation; Deep learning; Generative adversarial network (GAN)

Štítky

143160, CF PLANT, rivok

Příznaky

Mezinárodní význam, Recenzováno

Změněno: 24. 3. 2026 09:40, Mgr. Eva Dubská

Anotace

V originále

The generation of a large amount of ground truth data is an essential bottleneck for the application of deep learning-based approaches to plant image analysis. In particular, the generation of accurately labeled images of various plant types at different developmental stages from multiple renderings is a laborious task that substantially extends the time required for AI model development and adaptation to new data. Here, generative adversarial networks (GANs) can potentially offer a solution by enabling widely automated synthesis of realistic images of plant and background structures. In this study, we present a two-stage GAN-based approach to generation of pairs of RGB and binary-segmented images of greenhouse-grown plant shoots. In the first stage, FastGAN is applied to augment original RGB images of greenhouse-grown plants using intensity and texture transformations. The augmented data were then employed as additional test sets for a Pix2Pix model trained on a limited set of 2D RGB images and their corresponding binary ground truth segmentation. This two-step approach was evaluated on unseen images of different greenhouse-grown plants. Our experimental results show that the accuracy of GAN predicted binary segmentation ranges between 0.88 and 0.95 in terms of the Dice coefficient. Among several loss functions tested, Sigmoid Loss enables the most efficient model convergence during the training achieving the highest average Dice Coefficient scores of 0.94 and 0.95 for Arabidopsis and maize images. This underscores the advantages of employing tailored loss functions for the optimization of model performance.

Návaznosti

EH22_008/0004581, projekt VaV

Název: Nové poznatky pro plodiny nové generace

Citovat

ULLAH, Sajid; Narendra NARISETTI; Kerstin NEUMANN; Thomas ALTMANN; Jan HEJÁTKO a Evgeny GLADILIN. Automated generation of ground truth images of greenhouse-grown plant shoots using a GAN approach. PLANT METHODS. LONDON: BMC, 2025, roč. 21, č. 1, s. 1-14. ISSN 1746-4811. Dostupné z: https://doi.org/10.1186/s13007-025-01441-1.

@article{2547271,
   author = {Ullah, Sajid and Narisetti, Narendra and Neumann, Kerstin and Altmann, Thomas and Hejátko, Jan and Gladilin, Evgeny},
   article_location = {LONDON},
   article_number = {1},
   doi = {https://doi.org/10.1186/s13007-025-01441-1},
   keywords = {High-throughput greenhouse imaging; Plant phenotyping; Image segmentation; Ground truth data generation; Deep learning; Generative adversarial network (GAN)},
   language = {eng},
   issn = {1746-4811},
   journal = {PLANT METHODS},
   title = {Automated generation of ground truth images of greenhouse-grown plant shoots using a GAN approach},
   url = {https://link.springer.com/article/10.1186/s13007-025-01441-1},
   volume = {21},
   year = {2025}
}

TY  - JOUR
ID  - 2547271
AU  - Ullah, Sajid - Narisetti, Narendra - Neumann, Kerstin - Altmann, Thomas - Hejátko, Jan - Gladilin, Evgeny
PY  - 2025
TI  - Automated generation of ground truth images of greenhouse-grown plant shoots using a GAN approach
JF  - PLANT METHODS
VL  - 21
IS  - 1
SP  - 1-14
EP  - 1-14
PB  - BMC
SN  - 17464811
KW  - High-throughput greenhouse imaging
KW  - Plant phenotyping
KW  - Image segmentation
KW  - Ground truth data generation
KW  - Deep learning
KW  - Generative adversarial network (GAN)
UR  - https://link.springer.com/article/10.1186/s13007-025-01441-1
N2  - The generation of a large amount of ground truth data is an essential bottleneck for the application of deep learning-based approaches to plant image analysis. In particular, the generation of accurately labeled images of various plant types at different developmental stages from multiple renderings is a laborious task that substantially extends the time required for AI model development and adaptation to new data. Here, generative adversarial networks (GANs) can potentially offer a solution by enabling widely automated synthesis of realistic images of plant and background structures. In this study, we present a two-stage GAN-based approach to generation of pairs of RGB and binary-segmented images of greenhouse-grown plant shoots. In the first stage, FastGAN is applied to augment original RGB images of greenhouse-grown plants using intensity and texture transformations. The augmented data were then employed as additional test sets for a Pix2Pix model trained on a limited set of 2D RGB images and their corresponding binary ground truth segmentation. This two-step approach was evaluated on unseen images of different greenhouse-grown plants. Our experimental results show that the accuracy of GAN predicted binary segmentation ranges between 0.88 and 0.95 in terms of the Dice coefficient. Among several loss functions tested, Sigmoid Loss enables the most efficient model convergence during the training achieving the highest average Dice Coefficient scores of 0.94 and 0.95 for Arabidopsis and maize images. This underscores the advantages of employing tailored loss functions for the optimization of model performance.
ER  -

ULLAH, Sajid; Narendra NARISETTI; Kerstin NEUMANN; Thomas ALTMANN; Jan HEJÁTKO a Evgeny GLADILIN. Automated generation of ground truth images of greenhouse-grown plant shoots using a GAN approach. \textit{PLANT METHODS}. LONDON: BMC, 2025, roč.~21, č.~1, s.~1-14. ISSN~1746-4811. Dostupné z: https://doi.org/10.1186/s13007-025-01441-1.

Přehled o publikaci