Two-Stage Multi-Crop Disease Detection Framework

Input

Preprocessing

Stage 1

Stage 2

Output

Image PreprocessingRotation, Flipping, Color jitter

Training Only:• Rotation• Flipping• Color jitter

• Resize to 224x224• RGB Normalization• Tensor conversion

Stage 1: Crop IdentificationCNN Models: DenseNet121, ResNet50, VGG16,MobileNetV2, EfficientNetB3

Stage 2: Disease Detection (Binary Classification)CNN Models: DenseNet121, ResNet50, VGG16,MobileNetV2, EfficientNetB3

Final Prediction & Diagnosis

new