A Multi-Scale Feature Focus and Dynamic Sampling-Based Model for Hemerocallis fulva Leaf Disease Detection

Agriculture, Journal Year: 2025, Volume and Issue: 15(3), P. 262 - 262

Published: Jan. 25, 2025

Hemerocallis fulva, essential to urban ecosystems and landscape design, faces challenges in disease detection due limited data reduced accuracy complex backgrounds. To address these issues, the fulva leaf dataset (HFLD-Dataset) is introduced, alongside Multi-Scale Enhanced Network (HF-MSENet), an efficient model designed improve multi-scale reduce misdetections. The Channel–Spatial Module (CSMSM) enhances localization capture of critical features, overcoming limitations feature extraction caused by inadequate attention characteristics. C3_EMSCP module improves fusion combining convolutional kernels group convolution, increasing adaptability interaction across scales. interpolation errors boundary blurring upsampling, DySample adapts sampling positions using a dynamic offset learning mechanism. This, combined with pixel reordering grid techniques, reduces preserves edge details. Experimental results show that HF-MSENet achieves mAP@50 mAP%50–95 scores 94.9% 80.3%, respectively, outperforming baseline 1.8% 6.5%. Compared other models, demonstrates significant advantages efficiency robustness, offering reliable support for precise fulva.

Language: Английский

---

A Multimodal Data Fusion and Embedding Attention Mechanism-Based Method for Eggplant Disease Detection

Plants, Journal Year: 2025, Volume and Issue: 14(5), P. 786 - 786

Published: March 4, 2025

A novel eggplant disease detection method based on multimodal data fusion and attention mechanisms is proposed in this study, aimed at improving both the accuracy robustness of detection. The integrates image sensor data, optimizing features through an embedded mechanism, which enhances model’s ability to focus disease-related features. Experimental results demonstrate that excels across various evaluation metrics, achieving a precision 0.94, recall 0.90, 0.92, mAP@75 0.91, indicating excellent classification object localization capability. Further experiments, ablation studies, evaluated impact different loss functions model performance, all showed superior performance for approach. combined with mechanism effectively model, making it highly suitable complex identification tasks demonstrating significant potential widespread application.

Language: Английский

---

YOLOv8-GDCI: Research on the Phytophthora Blight Detection Method of Different Parts of Chili Based on Improved YOLOv8 Model

Agronomy, Journal Year: 2024, Volume and Issue: 14(11), P. 2734 - 2734

Published: Nov. 20, 2024

Smart farms are crucial in modern agriculture, but current object detection algorithms cannot detect chili Phytophthora blight accurately. To solve this, we introduced the YOLOv8-GDCI model, which can disease on leaves, fruits, and stem bifurcations. The model uses RepGFPN for feature fusion, Dysample upsampling accuracy, CA attention capture, Inner-MPDIoU loss small detection. In addition, also created a dataset of bifurcations, conducted comparative experiments. results manifest that demonstrates outstanding performance across gamut comprehensive indicators. comparison with YOLOv8n an improvement 0.9% precision, increase 1.8% recall, remarkable enhancement 1.7% average precision. Although FPS decreases slightly, it still exceeds industry standard real-time (FPS > 60), thus meeting requirements

Language: Английский

---

ECVNet: A Fusion Network of Efficient Convolutional Neural Networks and Visual Transformers for Tomato Leaf Disease Identification

Agronomy, Journal Year: 2024, Volume and Issue: 14(12), P. 2985 - 2985

Published: Dec. 15, 2024

Tomato leaf diseases pose a significant threat to plant growth and productivity, necessitating the accurate identification timely management of these issues. Existing models for tomato disease recognition can primarily be categorized into Convolutional Neural Networks (CNNs) Visual Transformers (VTs). While CNNs excel in local feature extraction, they struggle with global recognition; conversely, VTs are advantageous extraction but less effective at capturing features. This discrepancy hampers performance improvement both model types task identification. Currently, fusion that combine still relatively scarce. We developed an efficient network named ECVNet recognition. Specifically, we first designed Channel Attention Residual module (CAR module) focus on channel features enhance model’s sensitivity importance channels. Next, created Fusion (CAF effectively extract integrate features, thereby improving spatial capabilities. conducted extensive experiments using Plant Village dataset AI Challenger 2018 dataset, achieving state-of-the-art cases. Under condition 100 epochs, achieved accuracy 98.88% 86.04% dataset. The introduction provides solution diseases.

Language: Английский

---