Feeding Behavior Quantification and Recognition for Intelligent Fish Farming Application:A Review

Applied Animal Behaviour Science, Journal Year: 2025, Volume and Issue: 285, P. 106588 - 106588

Published: March 5, 2025

Language: Английский

---

A novel multiscale feature enhancement network using learnable density map for red clustered pepper yield estimation

Frontiers in Plant Science, Journal Year: 2025, Volume and Issue: 16

Published: April 7, 2025

Accurate and automated yield estimation for red cluster pepper (RCP) is essential to optimise field management resource allocation. Traditional object detection-based methods often suffer from time-consuming labour-intensive annotation processes, as well suboptimal accuracy in dense environments. To address these challenges, this paper proposes a novel multiscale feature enhancement network (MFEN) that integrates learnable density map (LDM) accurate RCP estimation. The proposed method mainly involves three key steps. First, the kernel-based (KDM) was improved by integrating Swin Transformer (ST), resulting LDM method, which produces higher quality maps. Then, MFEN developed improve extraction This combines dilation convolution, residual structures, an attention mechanism effectively extract features. Finally, were jointly trained estimate both maps RCP. model achieved superior using conjunction with joint training. Firstly, integration of significantly model, 0.98% improvement over previous iteration. Compared other networks, had lowest mean absolute error (MAE) 5.42, root square (RMSE) 10.37 symmetric percentage (SMAPE) 11.64%. It also highest R-squared (R²) value 0.9802 on test dataset, beating best performing DSNet 0.98%. Notably, despite its multi-column structure, has significant advantage terms parameters, only 13.08M parameters (a reduction 3.18M compared classic single-column CSRNet). highlights model's ability achieve while maintaining efficient deployment capabilities. provides robust algorithmic support intelligent

Language: Английский

---

Improved you only look once for weed detection in soybean field under complex background

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 151, P. 110762 - 110762

Published: April 8, 2025

Language: Английский

---

A dual-branch model combining convolution and vision transformer for crop disease classification

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(4), P. e0321753 - e0321753

Published: April 24, 2025

Computer vision holds tremendous potential in crop disease classification, but the complex texture and shape characteristics of diseases make classification challenging. To address these issues, this paper proposes a dual-branch model for which combines Convolutional Neural Network (CNN) with Vision Transformer (ViT). Here, convolutional branch is utilized to capture local features while handle global features. A learnable parameter used achieve linear weighted fusion two types An Aggregated Local Perceptive Feed Forward Layer (ALP-FFN) introduced enhance model’s representation capability by introducing locality into encoder. Furthermore, constructs lightweight block using ALP-FFN self-attention mechanism reduce parameters computational cost. The proposed achieves an exceptional accuracy 99.71% on PlantVillage dataset only 4.9M 0.62G FLOPs, surpassing state-of-the-art TNT-S (accuracy: 99.11%, parameters: 23.31M, FLOPs: 4.85G) 0.6%. On Potato Leaf dataset, attains 98.78% accuracy, outperforming advanced ResNet-18 98.05%, 11.18M, 1.82G) 0.73%. effectively advantages CNN ViT maintaining design, providing effective method precise identification diseases.

Language: Английский

---

Intelligent Detection Method for Surface Defects of Particleboard Based on Super-Resolution Reconstruction

Forests, Journal Year: 2024, Volume and Issue: 15(12), P. 2196 - 2196

Published: Dec. 13, 2024

To improve the intelligence level of particleboard inspection lines, machine vision and artificial technologies are combined to replace manual with automatic detection. Aiming at problem missed detection false on small defects due large surface width, complex texture different defect shapes particleboard, this paper introduces image super-resolution technology proposes a reconstruction model for images. Based Transformer network, incorporates an improved SRResNet (Super-Resolution Residual Network) backbone network in deep feature extraction module extract information. The shallow features extracted by conv 3 × then fused Transformer, considering both local global This enhances quality makes details clearer. Through comparison traditional bicubic B-spline interpolation method, ESRGAN (Enhanced Super-Resolution Generative Adversarial Network), SwinIR (Image Restoration Using Swin Transformer), effectiveness is verified using objective evaluation metrics including PSNR, SSIM, LPIPS, demonstrating its ability produce higher-quality images more better visual characteristics. Finally, YOLOv8 compare rates between low-resolution images, mAP can reach 96.5%, which 25.6% higher than recognition rate.

Language: Английский

---

Load Forecasting for Commercial Buildings Using BiLSTM–Transformer Network and Cyber–Physical Cognitive Control Systems

Symmetry, Journal Year: 2024, Volume and Issue: 16(12), P. 1601 - 1601

Published: Nov. 30, 2024

With the widespread adoption of electric vehicles (EVs), their charging and discharging schedules pose new challenges for real-time load forecasting in commercial buildings. This study proposes a prediction model based on integration bidirectional long short-term memory (BiLSTM) networks Transformer architecture, along with introduction cognitive control system cyber–physical systems (CPS) to address issues such as data loss excessive computation time during process. The BiLSTM–Transformer significantly improves load-forecasting accuracy performance by combining time-series modeling global feature extraction capabilities. Additionally, includes user-aware (UACC) Microgrid Control Center Cognitive (MACC). UACC quantifies information gaps real adaptively adjusts strategies communication instability, while MACC employs Q-learning algorithms evaluate impact scheduling optimize power resource allocation. synergy between these mechanisms ensures stability predictive scenarios involving or disruptions. Experimental results demonstrate that achieves outstanding under complete conditions reduces errors loss, validating its superior robustness. provides reliable support

Language: Английский

---