Related Experiment Video
Updated: Jan 11, 2026
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
2.0K
A comparative deep learning methodology for plant insect image classification: Assessment of CNN architectures and
Md Tomal Ahmed Sajib1, Nazmul Huda Badhon1, Imrus Salehin1,2
1Department of Computer Science and Engineering, Daffodil International University, Dhaka 1216, Bangladesh.
Methodsx
|November 17, 2025
Summary
Deep learning models effectively identify insects for crop protection using the BAU-Insectv2 dataset. Specific data augmentation and optimizer choices significantly enhance model accuracy, especially with limited agricultural data.
Area of Science:
- Agricultural Science
- Computer Science
- Machine Learning
Background:
- Deep learning excels in agricultural image analysis for pest recognition.
- Effective pest identification is crucial for crop protection and yield optimization.
Purpose of the Study:
- To present a comparative methodology for plant-insect image classification using the BAU-Insectv2 dataset.
- To investigate the impact of data augmentation and optimizers on model performance with small, field-collected data.
Main Methods:
- Evaluated four convolutional neural network (CNN) architectures (ResNet101V2, EfficientNet-B1, InceptionV3, InceptionResNetV1) using transfer learning.
- Assessed six single-factor data augmentations and three optimizers (Adam, SGD, RMSprop).
- Performance measured using accuracy, precision, recall, and F1-score.
Main Results:
- Adam optimizer generally yielded the most stable high accuracy on limited data.
- Specific model-augmentation pairings demonstrated significant impact; EfficientNet-B1 with cropping achieved near-perfect accuracy.
- ResNet101V2 with rotation and InceptionV3 with brightness also showed competitive performance.
Conclusions:
- The study provides a reproducible pipeline and augmentation-aware guidance for insect recognition with scarce data.
- This approach enables robust insect identification for improved agronomic decision support.
- Optimizing data augmentation and optimizer selection is critical for deep learning in agricultural image analysis.
