Over the past decade, deep learning has transformed both image processing and computational chemistry, the twin pillars of imaging science and technology. This review brings together the three generative and representation-learning techniques that have driven this change generative adversarial networks (GANs), diffusion models, and transformers alongside the learning-based methods that have reshaped computational chemistry. We position each technique against its classical counterpart, examine its motivation and limitations, and review recent applications ranging from super-resolution, restoration, and synthesis to property prediction and generative inverse design. We then discuss the methodological and computational extensions needed for trustworthy research, and close by relating these developments to the solar-cell, battery, and polymer studies long featured in imaging science and technology research.