terclim by ICS banner
IVES 9 IVES Conference Series 9 A novel dataset and deep learning object detection benchmark for grapevine pest surveillance

A novel dataset and deep learning object detection benchmark for grapevine pest surveillance

Abstract

Flavescence dorée (FD) stands out as a significant grapevine disease with severe implications for vineyards. The American grapevine leafhopper (Scaphoideus titanus) serves as the primary vector, transmitting the pathogen that causes yield losses and elevated costs linked to uprooting and replanting. Another potential vector of FD is the mosaic leafhopper, Orientus ishidae, commonly found in agroecosystems. The current monitoring approach involves periodic human identification of chromotropic traps, a labor-intensive and time-consuming process.

Therefore, there is a compelling need to develop an automatic pest detection system, leveraging the recent progress in computer vision and deep learning techniques. However, the current progress in developing such a system is hindered by the lack of effective datasets to serve as ground-truth data for the training process.

To fill this gap, our study contributes a fully annotated dataset of S. titanus and Or. ishidae from yellow sticky traps. The dataset comprises more than 400 images, with 1000 identification per class. Guided by entomologists, the annotation task involved defining bounding boxes around relevant insects with corresponding class labels.

We trained and compared the performance of state-of-the-art object detection algorithms (YOLOv8 and Faster R-CNN). Pre-processing included automatic cropping to eliminate irrelevant background information and image enhancements to improve overall quality. Additionally, we tested the impact of altering image resolution, data augmentation, and single-class detection. Preliminary results achieved a high detection accuracy, with mAP@50 and F1-score above 90%, and mAP@50-95 around 70%, allowing a first deployment as an automatic annotation support tool.

DOI:

Publication date: June 14, 2024

Issue: Open GPB 2024

Type: Poster

Authors

Giorgio Checola1*, Paolo Sonego1, Valerio Mazzoni2, Franca Ghidoni3, Alberto Gelmetti3, Pietro Franceschi1

1 Research and Innovation Centre, Digital Agriculture Unit, Fondazione Edmund Mach, S. Michele all’Adige, TN, Italy
2 Research and Innovation Centre, Plant Protection Unit, Fondazione Edmund Mach, S. Michele all’Adige, TN, Italy
3 Technology Transfer Centre, Viticulture Unit, Fondazione Edmund Mach, S. Michele all’Adige, TN, Italy

Contact the author*

Keywords

insect detection, deep learning, smart pest monitoring, flavescence dorée, insect traps

Tags

IVES Conference Series | Open GPB | Open GPB 2024

Citation

Related articles…

Development of a new method to understand headspace aroma distribution and explore the pre-sensory level in perceptive interactions involved in red wine fruity aroma expression

A part, at least, of red wines fruity expression may be explained by perceptive interactions involving particularly various substituted ethyl esters and acetates present at concentration far below their olfactory threshold, specifically thanks to synergistic effects. Wine sensory perception is directly linked to the stimulation of the taster at the level of olfactory epithelium by volatiles. These compounds are liberated from the matrix to the atmosphere, and will then be smelt. From a physico-chemical point of view, these volatiles ability to be released may be evaluated by their partition coefficients, which correspond to the volatile concentration ratio between the liquid and gas phase. Our goal is, through these coefficients determination, to assess if volatile matrix composition is able to impact the volatility of some compounds, and then explain sensory perception, i.eto evaluate what is called the pre-sensorial level impact.

Vine phenology and climate in Bordeaux, since the beginning of the XIXth century

We analyze the effects of climate (temperature and pluviometry) on the phenologic stages of the vine (débourrement, flowering, ripening and grape harvest). We rebuilt time series starting from the beginning of the XIXth century for the Medoc and the area of Bordeaux, data very seldom mobilized by researchers.

Deep learning based models for grapevine phenology

the phenological evolution is a crucial aspect of grapevine growth and development. Accurate detection of phenological stages can improve vineyard management, leading to better crop yield and quality traits. However, traditional methods of phenological tracking such as on-site observations are time-consuming and labour-intensive. This work proposes a scalable data-driven method to automatically detect key phenological stages of grapevines using satellite data. Our approach applies to vast areas because it solely relies on open and satellite data having global coverage without requiring any in-field data from weather stations or other sensors making the approach extensible to other areas.

Effect of soil particle size on vine water status, leaf ABA content and berry quality in Nebbiolo grapes

The root and shoot abscisic acid (ABA) accumulation in response to water deficit and its relation with stomatal conductance is longtime known in grapevine. ABA-dependent and ABA-independent signalling response to osmotic stress coexist in sessile plants. In grapevine, the signaling role of ABA in response to water stress conditions and its influence on berry quality is critical to manage grapevine acclimation to climate change.

Viticultural practices: past, present and future

Practices in viticulture have greatly evolved in the last five decades. There were three objectives: improvement in the quality of the products, reduction in the production costs through mechanization