Due to the automatic capture of the gastrointestinal tract, capsule endoscopy has major drawbacks for physicians such as the long reading time and the risk of overlooking abnormalities. A computer-aided supporting system that automatically detects abnormalities would be extremely helpful, but a satisfactory system had not been developed before the appearance of convolutional neural networks (CNNs) methodology. A CNN is a type of artificial intelligence that surpasses human ability in image recognition. This chapter introduces a variety of state-of-the-art CNN-based systems for resolving the abovementioned problems in capsule endoscopy reading. Most of them are in the development phase, and the evaluation under real-life clinical scenarios is crucial for clinical implementation. We must consider how to use them and whether they can be trusted.