Transforming Agriculture with AI: Automated Crop Insect Identification Using InceptionV3 Transfer Learning

• Published in: 2023 First International Conference on Advances in Electrical, Electronics and Computational Intelligence (ICAEECI) pp. 1 - 6
• Main Authors: K, Thyagharajan. K., G, Kalaiarasi, G, Nirmala, P, Saravanan, T, Vignesh, V, Makashkumar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 19-10-2023
• Subjects: Accuracy; Agriculture; Crops; Data augmentation; Deep Learning; Feature extraction; InceptionV3; Insect Detection; Insects; Productivity; Smart Agriculture; Task analysis; Testing; Transfer learning
• DOI: 10.1109/ICAEECI58247.2023.10370844

Identifying crop insects and classifying them is a challenging task for agriculturalists. A transfer learning method is suggested in this paper to address this issue. The proposed approach uses InceptionV3 as a base deep learning model and it is given additional custom layers. The model is fine-tuned using a custom dataset made up of five classes with 1000 insects each. The insect and pest images available from the National Bureau of Agricultural Insect Resources datasets and the Xie1 dataset are used for testing the model. Preprocessing techniques are applied to resize the images, and data augmentation is employed to prevent overfitting. The model was analyzed on the validation dataset with Adam, Adagrad, and Nadam optimizers, and obtained recognition accuracies of 86.6%, 77.8%, and 88.56% respectively. The outcomes confirms that the model performs well with Nadam optimizer for insect identification.