We explore the development of a set of algorithms for accurately localizing and classifying handwritten digits, addressing the challenges posed by variations in individual writing styles, digit sizes, and the presence of multiple digits in an image. Our work is structured into three tasks, each building upon the previous ones. Task A, which is trivial, focuses on classifying individual handwritten digits with 28 \(\times \) 28 pixel resolution using a convolutional neural network. Task B extends the CNN in Task A, as a black box, by developing algorithms to recognize multiple digits with identical dimension sizes, placed on a large image. Task C further complicates Task B by considering digits with varying dimension sizes in a large image. Our proposed approach involves the use of convolution operations for digit localization, and takes advantage of inspirations from existing algorithms for handling varying digit dimension sizes. Experimental results demonstrate the effectiveness of our approach in achieving high accuracy rates. Our work contributes to the advancement of robust algorithms capable of accurately classifying handwritten digits in different scenarios and is expected to be applicable to classifying other handwritten objects, such as handwritten letters.