Handwritten Digit Recognition of MNIST Dataset using Residual Network and Visual Geometry Group Network

In this work, two different deep learning architectures Residual Network (ResNet) and VGG Network are implemented for the MNIST digit recognition challenge. With a changed architecture, the ResNet which was initially created for challenging image identification tasks is tailored to the more straightforward MNIST dataset. Remaining blocks in the ResNet allow the network to learn hierarchical characteristics effectively. Furthermore, repeated 3x3 convolutional layers are used in an adaptation of the VGG architecture, which is well known for its simplicity and efficiency. Accuracy metrics and visualizations are used to assess the performance of both models after they have been trained and assessed on the MNIST dataset.With the potential for additional optimization and expansion to more complicated datasets, the code that is provided provides a fundamental framework for investigating and comprehending the use of ResNets and VGG Networks in the context of digit recognition task.