AgriCleanNet: A Sustainable Transformer-Guided Multi-Modal Framework with Explainable AI and Satellite-Aided Environmental Awareness for Indian Fruit Disease Detection

DOI

10.2991/978-94-6463-948-3_45How to use a DOI?

Keywords

Environmental Context; Explainable AI; Fruit Disease Detection; Geo-tagged Dataset; Multi-modal Learning; Precision Farming; Satellite Data; Smart Agriculture; Sustainable Agriculture; Vision Transformer

Abstract

Fruit diseases are becoming a bigger threat to farming in tropical areas. These diseases often go undetected until they have done a lot of damage. Traditional methods of finding diseases are slow, manual, and not very scalable. Most AI-based models only look at fruit images and don't take into account important environmental factors like humidity, rainfall, and temperature that help diseases spread. Also, a lot of models work like black boxes, which makes farmers and other stakeholders less likely to trust and use them. This paper proposes AgriCleanNet, an innovative multi-modal, transformer-guided, and explainable AI framework for fruit disease detection, motivated by the need for an interpretable, sustainable, and scalable solution. The main idea is to combine images of fruit, satellite-based environmental data, and agricultural context into one learning model. AgriCleanNet uses a Vision Transformer (ViT) backbone to process pictures of fruit, and a parallel encoder to process geo-tagged satellite-derived data like NDVI, rainfall, and humidity. A cross-modal attention fusion module facilitates contextual learning between visual and environmental modalities. To make the model more transparent, it now has Layer-wise Relevance Propagation (LRP) and Grad-CAM to explain its predictions. The work present GeoFruit++, a curated dataset that combines annotated images of fruit diseases with satellite and weather data from major Indian horticultural zones. This dataset will be used to train and test the model. Tests show that AgriCleanNet is 91.5% accurate, has a 91% F1-score, and is 15% more generalizable than the best baselines. The suggested method not only makes it easier to find diseases early on in different situations, but it also fits with the UN SDG goals of Zero Hunger and Climate Action, making it possible to use data-driven farming methods in real life.

Copyright

© 2025 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  - Rakesh Suryawanshi
AU  - Kailas Patil
PY  - 2026
DA  - 2026/01/06
TI  - AgriCleanNet: A Sustainable Transformer-Guided Multi-Modal Framework with Explainable AI and Satellite-Aided Environmental Awareness for Indian Fruit Disease Detection
BT  - Proceedings of the International Conference on Sustainable Innovation with Artificial Intelligence and Machine Learning 2025 (ICSIAIML 2025)
PB  - Atlantis Press
SP  - 641
EP  - 664
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6463-948-3_45
DO  - 10.2991/978-94-6463-948-3_45
ID  - Suryawanshi2026
ER  -