2 research outputs found
A novel dataset and deep learning object detection benchmark for grapevine pest surveillance
No full textFlavescence dorée (FD) poses a significant threat to grapevine health, with the American grapevine leafhopper, Scaphoideus titanus, serving as the primary vector. FD is responsible for yield losses and high production costs due to mandatory insecticide treatments, infected plant uprooting, and replanting. Another potential FD vector is the mosaic leafhopper, Orientus ishidae, commonly found in agroecosystems. The current monitoring approach, which involves periodic human identification of yellow sticky traps, is labor-intensive and time-consuming. Therefore, there is a compelling need to develop an automatic pest detection system leveraging recent advances in computer vision and deep learning techniques. However, progress in developing such a system has been hindered by the lack of effective datasets for training. To fill this gap, our study contributes a fully annotated dataset of S. titanus and O. ishidae from yellow sticky traps, which includes more than 600 images, with approximately 1500 identifications per class. Assisted by entomologists, we performed the annotation process, trained, and compared the performance of two state-of-the-art object detection algorithms: YOLOv8 and Faster R-CNN. Pre-processing, including automatic cropping to eliminate irrelevant background information and image enhancements to improve the overall quality of the dataset, was employed. Additionally, we tested the impact of altering image resolution and data augmentation, while also addressing potential issues related to class detection. The results, evaluated through 10-fold cross validation, revealed promising detection accuracy, with YOLOv8 achieving an [email protected] of 92%, and an F1-score above 90%, with an mAP@[0.5:0.95] of 66%. Meanwhile, Faster R-CNN reached an [email protected] and mAP@[0.5:0.95] of 86% and 55%, respectively. This outcome offers encouraging prospects for developing more effective management strategies in the fight against Flavescence dorée
Apple phenotyping using deep learning and 3D depth analysis: An experimental study on fruitlet sizing during early development
No full textCurrent research in apple-growing focuses on collecting extensive biometric data to better understand physiological processes, improve orchard productivity and predict yields. In this context, fruit thinning has emerged as a key horticultural practice to enhance fruit size and quality while preventing alternate bearing. Despite the growing role of plant imaging technologies in agronomic management, fruitlet sizing remains challenging, particularly in early phenological stages. To address this challenge, we developed an RGB-D-based vision pipeline that combines YOLO models with depth information and relies on the statistical analysis of frame series to detect and cluster fruitlets into flower corymbs, providing both fruitlet counting and diameter estimates for each video acquisition. After obtaining an [email protected] and AP@[0.5:0.95] of respectively 0.894 and 0.77 in fruitlet detection, along with a precision of 0.881 and a recall of 0.846, our approach efficiently processed video frames, extracting the most reliable data for each labeled cluster. While the comparison of true positive estimates with calibrated caliper measurements showed a mean RMSE of 1.05 mm, challenges remain in achieving the correct fruitlet count, with a mean counting error of 0.63 fruitlets per video. Additionally, the proposed workflow retrieved the exact number of fruitlets as the ground truth in 56.4% of the videos, increasing to 75% when excluding those videos where the correct fruitlet count was never detected in any frame by the YOLO model. Despite these limitations, our results are promising, proposing a potential data acquisition tool without compromising the reliability of traditional practices. This approach could pave the way for future applications, including the evaluation of plant growth regulator trials and the development of predictive models for yield and productivity optimization
