Toward Grapevine Digital Ampelometry Through Vision Deep Learning Models

Magalhães, Sandro Augusto; Castro, Luís; Rodrigues, Leandro; Padilha, Tiago Cerveira; Carvalho, Frederico De; Santos, Filipe Neves dos; Pinho, Tatiana M.; Moreira, Germano; Cunha, Jorge; Cunha, Mário; Silva, Paulo; Moreira, António Paulo

doi:10.1109/jsen.2023.3261544

Cited by 11 publications

(8 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…6C). If (Magalhães et al, 2023). They also 460 overcome problems associated with noise and leverage high numbers of de novo features.…”

Section: Resultsmentioning

confidence: 99%

“…We note that the vein and blade traces we produce on raw images are the perfect training set for automatic feature detection by machine learning models. Contrastingly, machine learning and convolutional neural networks can achieve high classification rates directly from images (Magalhães et al, 2023). They also overcome problems associated with noise and leverage high numbers of de novo features.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

A high resolution model of the grapevine leaf morphospace predicts synthetic leaves

Chitwood,

Torres-Lomas,

Hadi

et al. 2024

Preprint

View full text Add to dashboard Cite

Grapevine leaves are a model morphometric system. Sampling over ten thousand leaves using dozens of landmarks, the genetic, developmental, and environmental basis of leaf shape has been studied and a morphospace for the genus Vitis predicted. Yet, these representations of leaf shape fail to capture the exquisite features of leaves at high resolution. We measure the shapes of 139 grapevine leaves using 1672 pseudo-landmarks derived from 90 homologous landmarks with Procrustean approaches. From hand traces of the vasculature and blade, we have derived a method to automatically detect landmarks and place pseudo-landmarks that results in a high-resolution representation of grapevine leaf shape. Using polynomial models, we create continuous representations of leaf development in 10 Vitis spp. We visualize a high-resolution morphospace in which genetic and developmental sources of leaf shape variance are orthogonal to each other. Using classifiers, V. vinifera, Vitis spp., rootstock and dissected leaf varieties as well as developmental stages are accurately predicted. Theoretical eigenleaf representations sampled from across the morphospace that we call synthetic leaves can be classified using models. By predicting a high-resolution morphospace and delimiting the boundaries of leaf shapes that can plausibly be produced within the genus Vitis, we can sample synthetic leaves with realistic qualities. From an ampelographic perspective, larger numbers of leaves sampled at lower resolution can be projected onto this high-resolution space; or, synthetic leaves can be used to increase the robustness and accuracy of machine learning classifiers.

show abstract

“…6C). If (Magalhães et al, 2023). They also 460 overcome problems associated with noise and leverage high numbers of de novo features.…”

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

A high resolution model of the grapevine leaf morphospace predicts synthetic leaves

Chitwood,

Torres-Lomas,

Hadi

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…In addition, Magalhães et al [43] and Fuentes et al [36] were the only studies that were concerned with the position of the leaves used in the classification. Magalhães et al argue that leaves from nodes between the 7th and 8th should be used, as they are the most representative in terms of phenotypic characteristics [66], while Fuentes et al used mature leaves from the fifth position.…”

Section: Nomentioning

confidence: 99%

“…In fact, the ability of DL-based approaches to automatically learn to extract useful features has been making exponential advances in computer vision since 2012, with the publication of Krizhevsky et al [23]. Furthermore, the dataset of Koklu et al [7] is very small compared to the largest dataset used by DL-based architectures, Magalhães et al [43], which using small DL-based architectures managed to surpass the results achieved by Abassi and Jalal [30] (94.75 vs 83.20). Given this fact, the focus of the discussion and future directions in this study will be directed towards DL-based approaches.…”

Section: Machine Learning Vs Deep Learningmentioning

confidence: 99%

“…Focusing on the studies by Pereira et al [58] and Nasiri et al [53], they used the pre-trained AlexNet and VGG-16 models, respectively. Despite good results in the past, ImageNet classification results show that there is a high probability of improving results if more recent pre-trained CNN architectures are used, as in Carneiro et al [45], Magalhães et al [43], or Kunduracioglu and Pacal [38]. In addition, there are newer CNN architectures that remain untested, for example MobileNetV3 [118], larger members of the EfficientNet family [84,119], ConvNext [120,121].…”

Section: Pre-processingmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning and Machine Learning for Automatic Grapevine Varieties Identification: A Brief Review

Carneiro,

Cunha,

Sousa

2024

Preprint

View full text Add to dashboard Cite

The Eurasian grapevine (Vitis vinifera L.) is the most widely grown horticultural crop in the world and is important for the economy of many countries. In the wine production chain, grape varieties play an important role as they directly influence the authenticity and classification of the product. Identifying the different grape varieties is therefore fundamental for quality control and inspection activities, as well as for regulating production. Currently, ampelography and molecular analysis are the main approaches to identifying grape varieties. However, both methods have limitations. Ampelography is subjective and prone to errors and is experiencing enormous difficulties as ampelographers are increasingly scarce. On the other hand, molecular analyses are very demanding in terms of cost and time. In this scenario, Deep Learning (DL) and Machine Learning (ML) methods have emerged as a classification alternative to deal with the scarcity of ampelographs and avoid molecular analyses. In this study, the most recent and current methods for identifying grapevine varieties using DL classification-based approaches are presented through a systematic literature review. The classification pipeline of the 31 studies found in the literature was described, highlighting its pros and cons. Most of the studies used DL-based models trained with leaf images acquired in a controlled environment at a maximum distance of 1.2 metres to classify grape varieties. In addition, there is a large gap between practical applications and the datasets used: a great lack of varieties, limited data acquired in the field and a lack of tests on plants under adverse conditions. Potential directions for improving this area of research were also presented.

show abstract

A high‐resolution model of the grapevine leaf morphospace predicts synthetic leaves

Chitwood,

Torres‐Lomas,

Hadi

et al. 2024

Plants People Planet

View full text Add to dashboard Cite

Societal Impact StatementGrapevine leaves are emblematic of the strong visual associations people make with plants. Leaf shape is immediately recognizable at a glance, and therefore, this is used to distinguish grape varieties. In an era of computationally enabled machine learning‐derived representations of reality, we can revisit how we view and use the shapes and forms that plants display to understand our relationship with them. Using computational approaches combined with time‐honored methods, we can predict theoretical leaves that are possible, enabling us to understand the genetics, development, and environmental responses of plants in new ways.Summary Grapevine leaves are a model morphometric system. Sampling over 10,000 leaves using dozens of landmarks, the genetic, developmental, and environmental basis of leaf shape has been studied and a morphospace for the genus Vitis predicted. Yet, these representations of leaf shape fail to capture the exquisite features of leaves at high resolution. We measure the shapes of 139 grapevine leaves using 1672 pseudo‐landmarks derived from 90 homologous landmarks with Procrustean approaches. From hand traces of the vasculature and blade, we have derived a method to automatically detect landmarks and place pseudo‐landmarks that results in a high‐resolution representation of grapevine leaf shape. Using polynomial models, we create continuous representations of leaf development in 10 Vitis spp. We visualize a high‐resolution morphospace in which genetic and developmental sources of leaf shape variance are orthogonal to each other. Using classifiers, Vitis vinifera, Vitis spp., rootstock and dissected leaf varieties as well as developmental stages are accurately predicted. Theoretical eigenleaf representations sampled from across the morphospace that we call synthetic leaves can be classified using models. By predicting a high‐resolution morphospace and delimiting the boundaries of leaf shapes that can plausibly be produced within the genus Vitis, we can sample synthetic leaves with realistic qualities. From an ampelographic perspective, larger numbers of leaves sampled at lower resolution can be projected onto this high‐resolution space, or, synthetic leaves can be used to increase the robustness and accuracy of machine learning classifiers.

show abstract

Toward Grapevine Digital Ampelometry Through Vision Deep Learning Models

Cited by 11 publications

References 28 publications

A high resolution model of the grapevine leaf morphospace predicts synthetic leaves

A high resolution model of the grapevine leaf morphospace predicts synthetic leaves

Deep Learning and Machine Learning for Automatic Grapevine Varieties Identification: A Brief Review

A high‐resolution model of the grapevine leaf morphospace predicts synthetic leaves

Contact Info

Product

Resources

About