An AI Tomato Leaf Doctor Using MobileNetV2 and Streamlit: A Lightweight Deep Learning Tool for Farmers

Makepeace Zulu
DLI 2025 Research Track, PMLR 302:1-12, 2026.

Abstract

Tomato farming in Eswatini faces severe yield losses (up to 70 percent) due to diseases such as late blight (Phytophthora infestans) and bacterial leaf spot (Xanthomonas spp.), caused by climate variability and limited access to localized diagnostic tools. Existing AI solutions, such as Plantix, lack region-specific treatment advice, excluding farmers in low bandwidth areas. This study introduces AI Tomato Leaf Doctor, a Streamlit deployed, two stage MobileNetV2 system pretrained on ImageNet and fine-tuned. From experimental analysis, it is seen that plant leaf disease detection using MobileNetV2 outperforms existing approach in terms of accuracy as well as training time. The first model filters non tomato inputs, other plants or objects, while the second classifies ten tomato conditions (nine diseases and one healthy) using a PlantVillage dataset of 20,000 images (2000 per class). The model achieved 97 percent accuracy. The app provides chemical dosages for locally available fungicides and their market prices in Eswatini Lilangeni (SZL), delivered via a farmer friendly interface. By leveraging MobileNetV2’s lightweight design and Streamlit’s cloud compatibility, this work bridges the gap between AI innovation and the practical needs of Eswatini’s tomato-dependent households. Keywords: Tomato Disease Detection, MobileNetV2, Streamlit, PlantVillage Dataset, Eswatini Agriculture, Lightweight CNN.

Cite this Paper

BibTeX
@InProceedings{pmlr-v302-zulu26a, title = {An AI Tomato Leaf Doctor Using MobileNetV2 and Streamlit: A Lightweight Deep Learning Tool for Farmers}, author = {Zulu, Makepeace}, booktitle = {DLI 2025 Research Track}, pages = {1--12}, year = {2026}, editor = {Haddad, Hatem and Kahira, Albert Njoroge and Bourhim, Sofia and Olatunji, Iyiola Emmanuel and Makhafola, Lesego and Mwase, Christine}, volume = {302}, series = {Proceedings of Machine Learning Research}, month = {17--22 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v302/main/assets/zulu26a/zulu26a.pdf}, url = {https://proceedings.mlr.press/v302/zulu26a.html}, abstract = {Tomato farming in Eswatini faces severe yield losses (up to 70 percent) due to diseases such as late blight (Phytophthora infestans) and bacterial leaf spot (Xanthomonas spp.), caused by climate variability and limited access to localized diagnostic tools. Existing AI solutions, such as Plantix, lack region-specific treatment advice, excluding farmers in low bandwidth areas. This study introduces AI Tomato Leaf Doctor, a Streamlit deployed, two stage MobileNetV2 system pretrained on ImageNet and fine-tuned. From experimental analysis, it is seen that plant leaf disease detection using MobileNetV2 outperforms existing approach in terms of accuracy as well as training time. The first model filters non tomato inputs, other plants or objects, while the second classifies ten tomato conditions (nine diseases and one healthy) using a PlantVillage dataset of 20,000 images (2000 per class). The model achieved 97 percent accuracy. The app provides chemical dosages for locally available fungicides and their market prices in Eswatini Lilangeni (SZL), delivered via a farmer friendly interface. By leveraging MobileNetV2’s lightweight design and Streamlit’s cloud compatibility, this work bridges the gap between AI innovation and the practical needs of Eswatini’s tomato-dependent households. Keywords: Tomato Disease Detection, MobileNetV2, Streamlit, PlantVillage Dataset, Eswatini Agriculture, Lightweight CNN.} }
Endnote
%0 Conference Paper %T An AI Tomato Leaf Doctor Using MobileNetV2 and Streamlit: A Lightweight Deep Learning Tool for Farmers %A Makepeace Zulu %B DLI 2025 Research Track %C Proceedings of Machine Learning Research %D 2026 %E Hatem Haddad %E Albert Njoroge Kahira %E Sofia Bourhim %E Iyiola Emmanuel Olatunji %E Lesego Makhafola %E Christine Mwase %F pmlr-v302-zulu26a %I PMLR %P 1--12 %U https://proceedings.mlr.press/v302/zulu26a.html %V 302 %X Tomato farming in Eswatini faces severe yield losses (up to 70 percent) due to diseases such as late blight (Phytophthora infestans) and bacterial leaf spot (Xanthomonas spp.), caused by climate variability and limited access to localized diagnostic tools. Existing AI solutions, such as Plantix, lack region-specific treatment advice, excluding farmers in low bandwidth areas. This study introduces AI Tomato Leaf Doctor, a Streamlit deployed, two stage MobileNetV2 system pretrained on ImageNet and fine-tuned. From experimental analysis, it is seen that plant leaf disease detection using MobileNetV2 outperforms existing approach in terms of accuracy as well as training time. The first model filters non tomato inputs, other plants or objects, while the second classifies ten tomato conditions (nine diseases and one healthy) using a PlantVillage dataset of 20,000 images (2000 per class). The model achieved 97 percent accuracy. The app provides chemical dosages for locally available fungicides and their market prices in Eswatini Lilangeni (SZL), delivered via a farmer friendly interface. By leveraging MobileNetV2’s lightweight design and Streamlit’s cloud compatibility, this work bridges the gap between AI innovation and the practical needs of Eswatini’s tomato-dependent households. Keywords: Tomato Disease Detection, MobileNetV2, Streamlit, PlantVillage Dataset, Eswatini Agriculture, Lightweight CNN.
APA
Zulu, M.. (2026). An AI Tomato Leaf Doctor Using MobileNetV2 and Streamlit: A Lightweight Deep Learning Tool for Farmers. DLI 2025 Research Track, in Proceedings of Machine Learning Research 302:1-12 Available from https://proceedings.mlr.press/v302/zulu26a.html.

Related Material