A Lightweight Model for Weed Detection Based on the Improved YOLOv8s Network in Maize Fields DOI Creative Commons
Jinyong Huang, Xia Xu, Zhihua Diao

et al.

Agronomy, Journal Year: 2024, Volume and Issue: 14(12), P. 3062 - 3062

Published: Dec. 22, 2024

To address the issue of computational intensity and deployment difficulties associated with weed detection models, a lightweight target model for weeds based on YOLOv8s in maize fields was proposed this study. Firstly, network, designated as Dualconv High Performance GPU Net (D-PP-HGNet), constructed foundation (PP-HGNet) framework. introduced to reduce computation required achieve design. Furthermore, Adaptive Feature Aggregation Module (AFAM) Global Max Pooling were incorporated augment extraction salient features complex scenarios. Then, newly created network used reconstruct backbone. Secondly, four-stage inverted residual moving block (iRMB) employed construct iDEMA module, which replace original C2f feature module Neck improve performance accuracy. Finally, instead conventional convolution downsampling, further diminishing load. The new fully verified using established field dataset. test results showed that modified exhibited notable improvement compared YOLOv8s. Accuracy improved from 91.2% 95.8%, recall 87.9% 93.2%, [email protected] 90.8% 94.5%. number GFLOPs size reduced 12.7 G 9.1 MB, respectively, representing decrease 57.4% 59.2% model. Compared prevalent such Faster R-CNN, YOLOv5s, YOLOv8l, superior accuracy lightweight. paper effectively reduces cost hardware accurate identification limited resources.

Language: Английский

Multi-objective evolutionary co-learning framework for energy-efficient hybrid flow-shop scheduling problem with human-machine collaboration DOI
Jiawei Wu, Yong Liu, Yani Zhang

et al.

Swarm and Evolutionary Computation, Journal Year: 2025, Volume and Issue: 95, P. 101932 - 101932

Published: April 14, 2025

Language: Английский

Citations

0
Selective Scale-Aware Network for Traffic Density Estimation and Congestion Detection in ITS DOI Creative Commons
Jian Cheng, Chenxi Lin, Xiaojian Hu

et al.

Sensors, Journal Year: 2025, Volume and Issue: 25(3), P. 766 - 766

Published: Jan. 27, 2025

Traffic congestion detection in surveillance video is crucial for road traffic condition monitoring and improving operation efficiency. Currently, often characterized through density, which obtained by detecting vehicles or using holistic mapping methods. However, these traditional methods are not effective dealing with the vehicle scale variation video. This prompts us to explore density-map-based density Considering dynamic characteristics of flow, relying solely on spatial feature overly limiting. To address limitations, we propose a multi-task framework that simultaneously estimates congestion. Specifically, firstly Selective Scale-Aware Network (SSANet) generate map. Secondly, directly static level from map linear layer, can characterize occupancy each frame. In order further describe congestion, consider overall flow velocity integrated estimation assessment On collected dataset, our method achieves state-of-the-art results both task. SSANet also obtains 99.21% accuracy UCSD classification outperforms other

Language: Английский

Citations

0
Traffic congestion recognition based on convolutional neural networks in different scenarios DOI

Chao Wang,

Qiang Shang, Kun Liu

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 148, P. 110372 - 110372

Published: March 8, 2025

Language: Английский

Citations

0
GMMNet: A precise classification model for rice grains during rice processing DOI
J. Y. Liu, Jiahao Liu, Mingfang He

et al.

Expert Systems with Applications, Journal Year: 2025, Volume and Issue: 287, P. 128223 - 128223

Published: May 20, 2025

Language: Английский

Citations

0
A Lightweight Model for Weed Detection Based on the Improved YOLOv8s Network in Maize Fields DOI Creative Commons
Jinyong Huang, Xia Xu, Zhihua Diao

et al.

Agronomy, Journal Year: 2024, Volume and Issue: 14(12), P. 3062 - 3062

Published: Dec. 22, 2024

To address the issue of computational intensity and deployment difficulties associated with weed detection models, a lightweight target model for weeds based on YOLOv8s in maize fields was proposed this study. Firstly, network, designated as Dualconv High Performance GPU Net (D-PP-HGNet), constructed foundation (PP-HGNet) framework. introduced to reduce computation required achieve design. Furthermore, Adaptive Feature Aggregation Module (AFAM) Global Max Pooling were incorporated augment extraction salient features complex scenarios. Then, newly created network used reconstruct backbone. Secondly, four-stage inverted residual moving block (iRMB) employed construct iDEMA module, which replace original C2f feature module Neck improve performance accuracy. Finally, instead conventional convolution downsampling, further diminishing load. The new fully verified using established field dataset. test results showed that modified exhibited notable improvement compared YOLOv8s. Accuracy improved from 91.2% 95.8%, recall 87.9% 93.2%, [email protected] 90.8% 94.5%. number GFLOPs size reduced 12.7 G 9.1 MB, respectively, representing decrease 57.4% 59.2% model. Compared prevalent such Faster R-CNN, YOLOv5s, YOLOv8l, superior accuracy lightweight. paper effectively reduces cost hardware accurate identification limited resources.

Language: Английский

Citations

0