In this work, we explore the classification of tomato leaf diseases using deep learning techniques to identify whether leaves are healthy or diseased. Utilizing a custom dataset from PlantDoc and PlantVillage repositories, we implement and evaluate various convolutional neural network (CNN) architectures and transformers-based models. We also implement the LSGNet model, a lightweight neural network optimized for computational efficiency and real-time application. Our experiments involve feature extraction, fine-tuning, and deployment on a Raspberry Pi to assess the models’ practical performance. This project highlights the potential of integrating advanced AI techniques in agriculture, especially for early disease detection and crop health management. Furthermore, the deployment of these models on an IoT devices underscores their practical utility in enhancing farm operations and promoting sustainable farming practices.
Read more about the project here.
Datasets used for fine-tuning and features extraction:
- PlantVillage
- PlantDoc
Find more information about the datasets used and download the whole dataset from huggingface.co/tomato-leaves-dataset or direclty using the following prompt:
curl -X GET \
"https://datasets-server.huggingface.co/first-rows?dataset=lorenzoxi%2Ftomato-leaves-dataset&config=default&split=train"The following models were used for the experiments:
- ViT-Tiny-Patch16-224
- VGG16
- Swin-Tiny-Patch4-Window7-224
- ShuffleNetV2-x1.0
- ResNet50
- MobileViT-Small
- MobileNetV3-Large-100
- EfficientNetV2-RW-S
- AlexNet
- LSGNet
You can find and download already-trained models directly from here: download now.