Improving Rice Pest Management Through RP11: A Scientifically Annotated Dataset for Adult Insect Recognition DOI Creative Commons
Biao Ding, Yunxiang Tian,

Xiaojun Guo

et al.

Life, Journal Year: 2025, Volume and Issue: 15(6), P. 910 - 910

Published: June 4, 2025

Rice yields are expected to drop significantly due the increasing spread of rice pests. Detecting pests in a timely manner using deep learning models has become prevalent approach for rapid pest control. However, current datasets related often suffer from limited sample sizes or poorly annotated labels, which compromises training accuracy models. Building upon large-scale IP102 dataset, this study refines segment by separating adult specimens and larva specimens, acquiring additional images via web crawler techniques, re-annotating all samples. The category names, originally English, replaced with Latin scientific names corresponding families improve both clarity accuracy. resulting designated RP11, includes 11 categories 4559 7 larval 2467 images. All annotations follow labeling format compatible YOLO model training. count RP11 is approximately four times that rice-specific subset IP102. In work, YOLOv11 was employed evaluate RP11’s performance, serving as comparison dataset. results demonstrate outperforms precision (83.0% vs. 58.9%), recall (79.7% 63.1%), F1-score (81.3% 60.9%), mAP50 (87.2% 62.0%), mAP50–95 (73.3% 37.9%).

Language: Английский

Improving Rice Pest Management Through RP11: A Scientifically Annotated Dataset for Adult Insect Recognition DOI Creative Commons
Biao Ding, Yunxiang Tian,

Xiaojun Guo

et al.

Life, Journal Year: 2025, Volume and Issue: 15(6), P. 910 - 910

Published: June 4, 2025

Rice yields are expected to drop significantly due the increasing spread of rice pests. Detecting pests in a timely manner using deep learning models has become prevalent approach for rapid pest control. However, current datasets related often suffer from limited sample sizes or poorly annotated labels, which compromises training accuracy models. Building upon large-scale IP102 dataset, this study refines segment by separating adult specimens and larva specimens, acquiring additional images via web crawler techniques, re-annotating all samples. The category names, originally English, replaced with Latin scientific names corresponding families improve both clarity accuracy. resulting designated RP11, includes 11 categories 4559 7 larval 2467 images. All annotations follow labeling format compatible YOLO model training. count RP11 is approximately four times that rice-specific subset IP102. In work, YOLOv11 was employed evaluate RP11’s performance, serving as comparison dataset. results demonstrate outperforms precision (83.0% vs. 58.9%), recall (79.7% 63.1%), F1-score (81.3% 60.9%), mAP50 (87.2% 62.0%), mAP50–95 (73.3% 37.9%).

Language: Английский

Citations

0