A comparative analysis of convolutional neural networks approaches for phytoparasitic nematode identification

Abstract

Phytoparasitic nematode is a microscopic worm that affects the host plants and causes severe losses in the agricultural sector. Accurate and rapid identification of phytoparasitic nematodes is required to determine proper pest control and management. Hence it has become a necessity to automate the phytoparasitic nematode identification procedure. This study conducts a comparative analysis of 15 popular convolutional neural networks models, namely CoAtNet-0, DenseNet121, DenseNet169, DenseNet201, EfficientNetV2B0, EfficientNetV2B3, EfficientNetV2L, EfficientNetV2M, EfficientNetV2S, InceptionResNetV2, InceptionV3, ResNet101v2, ResNet50v2, VGG19, and Xception to deals with phytoparasitic nematode identification from the microscopic image. The results are compared using several evaluation metrics, namely test accuracy, mean class accuracy, F1 score, precision, and recall. The results show that CoAtNet-0 outperformed other models with 98.06% test accuracy, 97.86% mean class accuracy, 0.9803 F1 score, 0.9818 Precision, and 0.9806 Recall.