<![CDATA[Image classification using deep learning has become an effective approach in various fields, including visual object recognition such as food identification. This study aims to compare the performance of two well-known Convolutional Neural Network (CNN) architectures, AlexNet and ResNet50, in classifying traditional Indonesian food images. The dataset used in this research is a combination of two sources: a traditional Indonesian cake dataset from Kaggle and an additional set of images of Cirebon's traditional dishes. The final dataset consists of 24 food categories with more than 4,000 images in total. Each image was preprocessed through several steps including resizing to 224x224 pixels, applying data augmentation to training samples to enhance variation, and normalization based on standard input formats of the models. The training process was carried out using the 5-Fold Cross Validation method, while performance was evaluated using accuracy, precision, recall, and F1-score metrics. Experimental results show that ResNet50 consistently outperformed AlexNet across all evaluation metrics. ResNet50 achieved an average accuracy of 92%, compared to 86% obtained by AlexNet. Additionally, ResNet50 demonstrated superior performance in terms of precision, recall, and F1-score. This difference indicates that deeper and more complex architectures like ResNet50 are more effective in learning visual patterns in diverse traditional food images. The study concludes that ResNet50 is a more optimal choice for the task of traditional Indonesian food image classification. These findings serve as a basis for future development of image-based food recognition systems and support the preservation of culinary heritage through artificial intelligence technology.]]>
