Abstract:Accurate insect pest recognition plays a critical role in agriculture. It is a challenging problem due to the intricate characteristics of insects. In this paper, we present DeWi, novel learning assistance for insect pest classification. With a one-stage and alternating training strategy, DeWi simultaneously improves several Convolutional Neural Networks in two perspectives: discrimination (by optimizing a triplet margin loss in a supervised training manner) and generalization (via data augmentation). From that, DeWi can learn discriminative and in-depth features of insect pests (deep) yet still generalize well to a large number of insect categories (wide). Experimental results show that DeWi achieves the highest performances on two insect pest classification benchmarks (76.44\% accuracy on the IP102 dataset and 99.79\% accuracy on the D0 dataset, respectively). In addition, extensive evaluations and ablation studies are conducted to thoroughly investigate our DeWi and demonstrate its superiority. Our source code is available at <a class="link-external link-https" href="https://github.com/toannguyen1904/DeWi" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is **accurate identification of agricultural pests**, which is a crucial issue for agricultural production. Due to the complex characteristics of insects, such as morphological diversity, appearance changes at different growth stages, and similarities between species, pest identification is highly challenging. ### Specific description of the problem 1. **High intra - class differences and low inter - class differences**: Insects of the same species may have different morphologies at different growth stages (such as eggs, pupae, larvae), while insects of different species may be very similar visually. This makes pest identification a fine - grained classification problem that requires specific techniques to handle. 2. **Imbalanced data distribution**: The amount of data for some pest species is much larger than that for other species, resulting in the long - tail distribution problem. This imbalance will affect the generalization ability of the model. 3. **Small target and occlusion problems**: Pests are usually small and may be occluded by other objects (such as leaves, soil), which increases the difficulty of identification. ### Solutions proposed in the paper To solve the above problems, the paper proposes a new method named **DeWi (Deep - Wide Learning Assistance)** for insect pest classification. Specifically: - **Deep learning (Deep Step)**: By optimizing the triplet margin loss, DeWi extracts more discriminative features and enhances the discriminative ability of the model. The formula is as follows: \[ L_T(\{z^{(t)}, y^{(t)}\}_{t = 1}^B)=\sum_{t = 1}^B\left[m+\max_{p:y^{(p)}=y^{(t)}}D(z^{(t)}, z^{(p)})-\min_{n:y^{(n)}\neq y^{(t)}}D(z^{(t)}, z^{(n)})\right]_+ \] where \(m\) is the margin parameter, \(D(\cdot)\) is the distance function (this paper uses the Euclidean distance), and \([x]_+\) represents the maximum of \(x\) and 0. - **Generalized learning (Wide Step)**: Through the Mixup data augmentation technique, DeWi enhances the generalization ability of the model, enabling it to better adapt to a large number of different pest categories. The formulas of Mixup are as follows: \[ \tilde{x}=\lambda x^{(i)}+(1 - \lambda)x^{(j)} \] \[ \tilde{y}=\lambda y^{(i)}+(1 - \lambda)y^{(j)} \] where \(\lambda\sim\text{Beta}(\alpha,\alpha)\), \(\alpha\in(0,\infty)\). ### Experimental results The experimental results show that DeWi achieves the highest performance on both pest classification benchmark datasets IP102 and D0. For example, on the IP102 dataset, the accuracy of DeWi reaches 76.44%, and on the D0 dataset, the accuracy is as high as 99.79%. In conclusion, through proposing the DeWi method, this paper effectively solves multiple challenges in insect pest classification and significantly improves the classification performance.

Deep-Wide Learning Assistance for Insect Pest Classification

Effective Insect Recognition Using a Stacked Autoencoder with Maximum Correntropy Criterion

An Efficient Insect Pest Classification Using Multiple Convolutional Neural Network Based Models

Design of a Crop Disease and Pest Identification System Based on Deep Learning

Deep learning powered real-time identification of insects using citizen science data

An Efficient Deep Learning-based approach for Recognizing Agricultural Pests in the Wild

A two-stream network with complementary feature fusion for pest image classification

An Efficient Approach for Crops Pests Recognition and Classification Based on Novel DeepPestNet Deep Learning Model

Attention-Based Multiscale Feature Pyramid Network for Corn Pest Detection under Wild Environment

A lightweight model for efficient identification of plant diseases and pests based on deep learning

Classification Method of Significant Rice Pests Based on Deep Learning

A Crop Pests Image Classification Algorithm Based on Deep Convolutional Neural Network

Development of an Automatic Pest Monitoring System Using a Deep Learning Model of DPeNet

Classification and Morphological Analysis of Vector Mosquitoes using Deep Convolutional Neural Networks

Pest identification via deep residual learning in complex background

Based on the multi-scale information sharing network of fine-grained attention for agricultural pest detection

Common Pests Image Recognition Based on Deep Convolutional Neural Network

Causality-inspired crop pest recognition based on Decoupled Feature Learning

Deep learning techniques for enhancement of weeds growth classification

IoT-Enabled Pest Identification and Classification with New Meta-Heuristic-Based Deep Learning Framework

Deep Learning Based Automatic Multiclass Wild Pest Monitoring Approach Using Hybrid Global and Local Activated Features