<![CDATA[Handwritten digit recognition using the MNIST dataset is one of the applications in digital image processing. The selection of hyperparameters in the CNN architecture for handwriting recognition presents a challenge in achieving better recognition accuracy. This research focuses on the implementation of the DenseNet-201 architecture for recognizing handwritten digits in the MNIST dataset. The research stages include dataset preprocessing, model training, model testing, and model evaluation. The MNIST dataset consists of 60,000 training data and 10,000 testing data. Dataset preprocessing involves resizing the images to a larger size. The model training applies the DenseNet-201 architecture with selected hyperparameters such as activation functions (Softmax and ReLU), optimizers (Adam, RMSprop, and SGD), and learning rates (0.1, 0.01, and 0.001). The model testing uses one of the nine best-performing trained models. Model evaluation uses a confusion matrix to assess the accuracy and recognition performance on the MNIST dataset. The results show that the DenseNet-201 architecture with the RMSprop optimizer and a learning rate of 0.001 achieved a handwritten digit recognition accuracy of 99.49%. This study provides insights into CNN architectures and optimal hyperparameter selection for digital image processing]]>
Copyrights © 2025
