Detection of Apple Leaf Gray Spot Disease Based on Improved YOLOv8 Network

Mathematics, Journal Year: 2025, Volume and Issue: 13(5), P. 840 - 840

Published: March 3, 2025

In the realm of apple cultivation, efficient and real-time monitoring Gray Leaf Spot is foundation effective management pest control, reducing pesticide dependence easing burden on environment. Additionally, it promotes harmonious development agricultural economy ecological balance. However, due to dense foliage diverse lesion characteristics, disease faces unprecedented technical challenges. This paper proposes a detection model for apple, which based an enhanced YOLOv8 network. The details are as follows: (1) we introduce Dynamic Residual Blocks (DRBs) boost model’s ability extract features, thereby improving accuracy; (2) add Self-Balancing Attention Mechanism (SBAY) optimize feature fusion improve deal with complex backgrounds; (3) incorporate ultra-small head simplify computational reduce complexity network while maintaining high precision detection. experimental results show that outperforms original in detecting Spot. Notably, when Intersection over Union (IoU) 0.5, improvement 7.92% average observed. Therefore, this advanced technology holds pivotal significance advancing sustainable industry environment-friendly agriculture.

Language: Английский

ASE-YOLOv8n: A Method for Cherry Tomato Ripening Detection

Agronomy, Journal Year: 2025, Volume and Issue: 15(5), P. 1088 - 1088

Published: April 29, 2025

To enhance the efficiency of automatic cherry tomato harvesting in precision agriculture, an improved YOLOv8n algorithm was proposed for fast and accurate recognition natural environments. The improvements are as follows: first, ADown down-sampling module replaces part original network backbone’s standard convolution, enabling model to capture higher-level image features more target detection, while also reducing complexity by cutting number parameters. Secondly, model’s neck adopts a Slim-Neck (GSConv+VoV-GSCSP) instead traditional convolution with C2f. It this combination efficient CSConv swaps C2f VoV-GSCSP. Finally, introduces EMA attention mechanism, implemented at P5 layer, which enhances feature representation capability, extract detailed accurately. This study trained object-detection on self-built dataset before after improvement compared it early deep learning models YOLO series algorithms. experimental results show that increases accuracy 3.18%, recall 1.43%, F1 score 2.30%, mAP50 1.57%, mAP50-95 1.37%. Additionally, parameters is reduced 2.52 M, size 5.08 MB, outperforms other related previous version. experiment demonstrates technology’s broad potential embedded systems mobile devices. offers efficient, support automated harvesting.

Language: Английский

YOLOv8n-CSD: A Lightweight Detection Method for Nectarines in Complex Environments

Agronomy, Journal Year: 2024, Volume and Issue: 14(10), P. 2427 - 2427

Published: Oct. 19, 2024

At present, the picking of nectarines mainly relies on manual completion in China, and process involves high labor intensity during low efficiency. Therefore, it is necessary to introduce automated picking. To improve accuracy nectarine fruit recognition complex environments increase efficiency automatic orchard-picking robots, a lightweight detection method, YOLOv8n-CSD, proposed this study. This model improves YOLOv8n by first proposing new structure, C2f-PC, replace C2f structure used original network, thus reducing number parameters. Second, SEAM introduced model’s occluded part. Finally, realize real-time fruits, DySample Lightweight Dynamic Upsampling Module save computational resources while effectively enhancing anti-interference ability. With compact size 4.7 MB, achieves 95.1% precision, 84.9% recall, [email protected] 93.2%—the volume has been reduced evaluation metrics have all improved over baseline model. The study shows that YOLOv8n-CSD outperforms current mainstream target models, can recognize different faster more accurately, which lays foundation for field application technology.

Language: Английский