Indonesia has various types of spices used in culinary and traditional medicine. However, changes in lifestyle and modernization have made it increasingly difficult for the younger generation to recognize spices directly. Conventional identification still relies on manual observation which is prone to errors. Therefore, an artificial intelligence-based solution is needed to improve the accuracy of spice classification. This study applies the Convolutional Neural Network (CNN) method with the ResNet50 architecture, which is part of Deep Learning, to classify digital images of spices. This model utilizes Computer Vision to recognize visual patterns, Transfer learning to improve training efficiency, and Data Augmentation Techniques such as rotation, flipping, and scaling to improve model robustness. Evaluation using Confusion Matrix was carried out with various dataset division scenarios, including ratios of 90:10, 80:20, 70:30, 60:40, and 50:50. The experimental results showed that the model with a ratio of 90:10 provided the best accuracy, reaching 98.04%, with high precision, recall, and F1-score. In conclusion, the CNN method with ResNet50 has proven effective in identifying spices based on digital images. Further development can be done by adding variations of datasets and exploring other Deep Learning architectures to improve model performance.