On construction of data preprocessing for real-life SoyLeaf dataset & disease identification using Deep Learning...

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Deep learning models require extensive data for training, but high-quality soybean leaf disease datasets are scarce.
• This limitation hinders the development of accurate automated disease identification systems.

Purpose of the Study:

• To develop a comprehensive, real-life dataset for soybean leaf disease identification.
• To evaluate the performance of various pre-trained deep learning models for classifying soybean leaf diseases.

Main Methods:

• A new dataset, SoyLeaf, was created with 9786 high-quality images of healthy and diseased soybean leaves.
• Data preprocessing techniques were applied to enhance image quality.
• Fourteen Keras Transfer Learning models were fine-tuned and evaluated using Adam and RMSprop optimizers.

Main Results:

• Several fine-tuned models demonstrated high accuracy, exceeding 99% in many cases.
• MobileNetV2 and DenseNet121 achieved the highest accuracies of 99.89% with the Adam optimizer.
• ResNet50V2, ResNet101V2, InceptionV3, InceptionResNetV2, MobileNet, MobileNetV2, DenseNet121, and DenseNet169 showed superior performance.

Conclusions:

• The developed SoyLeaf dataset is a valuable resource for training deep learning models for soybean disease detection.
• Transfer learning approaches are highly effective for identifying soybean leaf diseases with high accuracy.
• Specific models like MobileNetV2 and DenseNet121 show exceptional promise for practical applications in agriculture.