Dilated-Windows-based Vision Transformer with Efficient-Suppressive-self-attention for insect pests classification

Hechen, Zhenzhe; Huang, Wei; Yin, Le; Xie, Wenjing; Zhao, Yixin

doi:10.1016/j.engappai.2023.107228

Cited by 3 publications

(1 citation statement)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Experiments on a dataset of 12 types of pests demonstrated the excellent performance of MFFNet, achieving a classification accuracy (ACC) of 98.2%. Hechen et al [23] The above research demonstrates the effectiveness of CNN models in pest classification. However, most studies have made few improvements to the model structure, and the used models do not have lightweight characteristics, resulting in low recognition efficiency.…”

Section: Introductionmentioning

confidence: 73%

A Lightweight Crop Pest Classification Method Based on Improved MobileNet-V2 Model

Peng,

Xu,

Shen

et al. 2024

Agronomy

View full text Add to dashboard Cite

This paper proposes PestNet, a lightweight method for classifying crop pests, which improves upon MobileNet-V2 to address the high model complexity and low classification accuracy commonly found in pest classification research. Firstly, the training phase employs the AdamW optimizer and mixup data augmentation techniques to enhance the model’s convergence and generalization capabilities. Secondly, the Adaptive Spatial Group-Wise Enhanced (ASGE) attention mechanism is introduced and integrated into the inverted residual blocks of the MobileNet-V2 model, boosting the model’s ability to extract both local and global pest information. Additionally, a dual-branch feature fusion module is developed using convolutional kernels of varying sizes to enhance classification performance for pests of different scales under real-world conditions. Lastly, the model’s activation function and overall architecture are optimized to reduce complexity. Experimental results on a proprietary pest dataset show that PestNet achieves classification accuracy and an F1 score of 87.62% and 86.90%, respectively, marking improvements of 4.20 percentage points and 5.86 percentage points over the baseline model. Moreover, PestNet’s parameter count and floating-point operations are reduced by 14.10% and 37.50%, respectively, compared to the baseline model. When compared with ResNet-50, MobileNet V3-Large, and EfficientNet-B1, PestNet offers superior parameter efficiency and floating-point operation requirements, as well as improved pest classification accuracy.

show abstract

Section: Introductionmentioning

confidence: 73%

A Lightweight Crop Pest Classification Method Based on Improved MobileNet-V2 Model

Peng,

Xu,

Shen

et al. 2024

Agronomy

View full text Add to dashboard Cite

show abstract

Pest-ConFormer: A hybrid CNN-Transformer architecture for large-scale multi-class crop pest recognition

Fang,

Tan,

Tang

et al. 2024

Expert Systems with Applications

View full text Add to dashboard Cite

Bugs and bytes: Entomological biomonitoring through the integration of deep learning and molecular analysis for merged community and network analysis

Suresh,

Xin,

Cook

et al. 2024

Agri and Forest Entomology

View full text Add to dashboard Cite

Insects play a vital role in ecosystem functioning, but in some parts of the world, their populations have declined significantly in recent decades due to environmental change, agricultural intensification and other anthropogenic drivers. Monitoring insect populations is crucial for understanding and mitigating biodiversity loss, especially in agro‐ecosystems where a focus on pests and beneficial insects is gaining momentum in the context of sustainable food systems. Biomonitoring has long played an important role in providing early warnings of insect pests and their vectored pathogens and for assessing agro‐ecosystem management. However, identification of invertebrates by taxonomists is time‐consuming and fraught with numerous other challenges, particularly when it comes to real‐time monitoring. Recent technological advances in both computational image recognition and molecular ecology have enhanced biomonitoring efficiency and accuracy, reducing labour efforts, but leading to unprecedented volumes of data generated. This perspective article examines the significance and further potential of deep learning with image‐based recognition, aided by complementary technologies, in advancing entomological biomonitoring. Using entomological sticky traps as an example, we discuss each step of the workflow required to create ecological networks using image‐based recognition, multimodal data and deep learning, and we identify the challenges inherent to this method and other insect survey techniques. In order to become mainstream for global biomonitoring, access to long‐term, standardised multimodal data is required for comprehending ecosystem dynamics, structure and function and for mitigating insect population declines.

show abstract

Dilated-Windows-based Vision Transformer with Efficient-Suppressive-self-attention for insect pests classification

Cited by 3 publications

References 25 publications

A Lightweight Crop Pest Classification Method Based on Improved MobileNet-V2 Model

A Lightweight Crop Pest Classification Method Based on Improved MobileNet-V2 Model

Pest-ConFormer: A hybrid CNN-Transformer architecture for large-scale multi-class crop pest recognition

Bugs and bytes: Entomological biomonitoring through the integration of deep learning and molecular analysis for merged community and network analysis

Contact Info

Product

Resources

About