MNIST handwritten digit recognition with different CNN architectures

Handwritten digit recognition has long been a popular research topic in computer vision and pattern recognition. Recognizing handwritten digits used to be challenging but thanks to many machine learning techniques nowadays, the problem is no longer. In this research, we looked into the MNIST database using fast.ai and trained the CNN ResNet-18 model to recognize handwritten digits. We then modified the architecture with different pre-trained models. For this work, we implemented five PyTorch’s pre-trained models, which are GoogLeNet, MobileNet v2, ResNet-50, ResNeXt-50, Wide ResNet-50. The purpose of this paper is to reveal the most accurate architecture for handwritten digits recognition. Also, we provide comparisons of training time, top-1 error, top-5 error and model size on all five models.