In recent years, deep learning has advanced biological image classification. This study investigates Convolutional Neural Network (CNN) architectures for classifying mushroom species and determining edibility. Using a dataset of 15,000 images across 25 classes, we developed a custom CNN and adapted AlexNet. Data was split 80% for training and 20% for validation. The custom CNN used multiple convolutional layers, ReLU activations, and max-pooling, optimized with the Adam optimizer and categorical cross-entropy loss. AlexNet’s deeper architecture was modified to suit our dataset. Both models incorporated dropout and early stopping to prevent overfitting. We evaluated performance using accuracy, precision, recall, and F1 score. Our custom CNN slightly outperformed AlexNet, demonstrating the potential of tailored architectures in biological classification. This research contributes to mushroom classification and supports the value of custom CNNs. Future work will expand the dataset and examine additional deep learning approaches.