Manuscript received July 25, 2025; revised August 13, 2025; accepted September 26, 2025; published February 27, 2026.

Abstract—Non-parasitic nematodes play a critical role in maintaining soil health and promoting crop productivity. Recognizing these roles is essential for developing effective and targeted strategies for sustainable farming. Valuable insights into the functional role of nematodes in soil can be obtained by categorizing them into bacterial feeders, fungal feeders, and the predatory-omnivorous group. However, traditional identification methods based on visual taxonomic traits are challenging and time-consuming. Advancements in artificial intelligence, particularly deep learning, have opened new frontiers in nematode identification; however, studies focusing on non-parasitic nematodes and their feeding behaviors remain limited. This study bridges this gap by leveraging state-of-the-art Convolutional Neural Network (CNN) architectures to classify non-parasitic nematodes from a small dataset of 921 microscopic photographs collected from agricultural fields and labeled according to their feeding habits. Ten CNN architectures, including Xception, VGG16, ResNet50, InceptionV3, InceptionResNetV2, NASNetMobile, DenseNet121, DenseNet201, EfficientNetV2B0, and ConvNeXtTiny, were evaluated for their performance. Among these, DenseNet121 emerged as the top performer, achieving a test accuracy of 0.86. These findings highlight the potential of deep learningbased techniques to revolutionize soil health management by enabling precise and efficient identification of nematodes.

Keywords—Convolutional Neural Networks (CNN), deep learning, non-parasitic nematode, precision agriculture, soil management

Cite: Nabila H. Shabrina, Larasati, Siwi Indarti, and Rina Maharani, "Convolutional Neural Networks for Non-parasitic Nematode Feeding Behavior Identification in Soil Ecosystem Management," Journal of Image and Graphics, Vol. 14, No. 1, pp. 108-121, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.