<![CDATA[Mushrooms are one of the most nutritious food sources; however, several species contain toxic compounds that can cause serious poisoning or even death if incorrectly identified. The high visual similarity between edible and poisonous mushrooms makes manual identification difficult, especially for non-experts. Therefore, an automatic image classification system is needed to improve the accuracy and consistency of mushroom identification. This study aims to develop a mushroom image classification model using a Convolutional Neural Network (CNN) optimized with the Adam Optimizer. The dataset was obtained from Kaggle and consisted of 2,820 images, divided into 2,256 training images, 282 validation images, and 282 testing images. The model was further validated using an external dataset of 83 real mushroom images to evaluate its generalization capability. All images were preprocessed through image resizing to 224 × 224 pixels and pixel normalization before model training. Model performance was evaluated using accuracy, precision, recall, F1-score, and Area Under the Curve (AUC). Experimental results on the Kaggle test dataset achieved an accuracy of 62.77%, precision of 58.28%, recall of 71.97%, F1-score of 64.41%, and an AUC of 0.6482. Evaluation on the external dataset demonstrated improved performance, achieving an accuracy of 85.54%, precision of 94.59%, recall of 77.78%, and F1-score of 85.37%. These findings indicate that the CNN model optimized with the Adam Optimizer is capable of performing mushroom image classification effectively and demonstrates good generalization performance on real-world data, making it a promising approach for automatic mushroom identification.]]>
