Using a two-stage convolutional neural network to rapidly identify tiny herbivorous beetles in the field
Preprint
Abstract
Recently, deep convolutional neural networks (CNN) have been adopted to help beginners identify insect species from field images. However, the application of these methods on the identification of tiny congeneric species moving across heterogeneous background remains difficult. To enable rapid and automatic identification in the field, we customized a method involving real-time object detection of two Phyllotreta beetles. We first performed data augmentation using transformations, syntheses, and random erasing of the original images. We then proposed a two-stage method for the detection and identification of small insects based on CNN, where YOLOv4 and EfficientNet were used as a region proposal network and a re-identification method, respectively. Evaluation of the model revealed that one-step object detection by YOLOv4 alone was not precise (Precision = 0.55) when classifying two species of flea beetles and background objects. In contrast, the two-step CNNs improved the precision (Precision = 0.89) with moderate accuracy (F-measure = 0.55) and acceptable speed (ca. 5 frames per second for full HD images) of detection and identification of insect species in the field. Although real-time identification of tiny insects remains a challenge in the field, our method aids in improving small object detection on a heterogeneous background.Keywords
All Related Versions
- Published version: Ecological Informatics, 66.
This publication has 34 references indexed in Scilit:
- Deep learning and computer vision will transform entomologyProceedings of the National Academy of Sciences, 2021
- Automatic identification for field butterflies by convolutional neural networksEngineering Science and Technology, an International Journal, 2020
- Species‐level image classification with convolutional neural network enables insect identification from habitus imagesEcology and Evolution, 2019
- Smart Insect CamerasBiodiversity Information Science and Standards, 2019
- Research on insect pest image detection and recognition based on bio-inspired methodsBiosystems Engineering, 2018
- Performance effect analysis for insect classification using convolutional neural network2017
- A survey on image-based insect classificationPattern Recognition, 2017
- SSD: Single Shot MultiBox Detector2016
- Deep learningNature, 2015
- Defence mechanisms of Brassicaceae: implications for plant-insect interactions and potential for integrated pest management. A reviewAgronomy for Sustainable Development, 2010