As an ML engineer at a manufacturing company, you are tasked with building a model to identify defects in products based on images taken at the end of the assembly line. The goal is to preprocess these images with lower computational requirements while efficiently extracting features related to defects. Which approach should you use to build the model?