A Standardised Multi-Source Dataset Construction and Comparative Evaluation of VGG-19, ResNet-50, and Inception V3 for Rice Disease Classification

DOI

10.2991/978-94-6239-789-7_32How to use a DOI?

Keywords

Rice disease detection; transfer learning; convolutional neural networks; dataset augmentation; image classification; precision agriculture

Abstract

The main staple meal is rice (Oryza sativa L.) of over 3.5 billion individuals worldwide, Asia contributing about 90 percent of the output. The impact of rice diseases on losses of annual crop is between 10 and 37 per cent, that posed threat to food security in developing countries. This paper has two main goals, (i) the creation of a standardised and balanced multi-source image dataset of 11,752 images that it has in seven categories Healthy, Brown Spot, Leaf Blast False Smut,, Sheath Blight, Neck Blast, and Bakkanae with systematic collection, preprocessing and augmentation protocols; and (ii) a stringent comparative performance testing of three commonly-used deep Experimental results highlight that Inception V3 achieves the highest classification accuracy of 96.15% (F1-score: 95.84%), followed by ResNet-50 at 94.72% (F1-score: 94.22%) and VGG-19 at 91.38% (F1-score: 90.72%). Class-wise analysis, confusion matrices, comparisons of computational efficiency and a standard with state-of-the-art methods are given. This research provides reliable dataset construction rules and model-selection criteria to agricultural disease detection tasks.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Kapil Kumar
AU  - Archana Sandhu
PY  - 2026
DA  - 2026/10/08
TI  - A Standardised Multi-Source Dataset Construction and Comparative Evaluation of VGG-19, ResNet-50, and Inception V3 for Rice Disease Classification
BT  - Proceedings of theInternational Conference on Innovation and Practices in Artificial Intelligence (ICIPAI 2026)
PB  - Atlantis Press
SP  - 471
EP  - 483
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-789-7_32
DO  - 10.2991/978-94-6239-789-7_32
ID  - Kumar2026
ER  -