Table 2.

Summary of the CNN architecture and learning parameters for machine learning models

Model	Architecture Details	Learning Parameters	Activation Function
CNN	Input: (224, 224, 1) → Conv2D (32, 3 × 3, ReLU) → MaxPooling (2 × 2) → Dropout (0.3) → Conv2D (64, 3 × 3, ReLU) → MaxPooling (2 × 2) → Dropout (0.3) → Conv2D (128, 3 × 3, ReLU) → MaxPooling (2 × 2) → Dropout (0.3) → Conv2D (256, 3 × 3, ReLU) → MaxPooling (2 × 2) → Dropout (0.3) → Conv2D (512, 3 × 3, ReLU) → MaxPooling (2 × 2) → Dropout (0.3) → Flatten → Dense (256, ReLU) → Dropout (0.3) → Dense (4, Softmax)	Optimizer: Adam Loss: Sparse Categorical Cross-Entropy Early Stopping: 10 Batch Size: 16 Epochs: 30 Dropout: 0.3–0.5 Batch Normalization: Yes Cross-Validation: 5-fold	ReLU (hidden) Softmax (output)
SVM	-	Standardize Data: True, Solver: SMO, Cross-Validation: 5, Kernel: RBF, C: 1.0, Gamma: ‘scale’, probability = True, Cross-Validation: 5	-
DT	-	Standardize Data: True, Criterion: ‘gini’, Max Depth: None, Min Samples Split: 2, Estimators: 100, Cross-Validation: 5,	-
RF	-	Standardize Data: True, Cross-Validation: 5, Number of Estimators: 100, Criterion: ‘gini’, Max Features: ‘auto’, Cross-Validation: 5	-