An Introductory Survey on Attention Mechanisms in Computer Vision Problems

Attention mechanism is a simple method derived from the research of human vision. With the development of artificial intelligence, it is gradually applied to the field of deep learning, especially computer vision. The weight of the pixels is redistributed in the form of a mask to compose attention mechanism. Various tasks in the field of computer vision, including image classification, object detection, image semantic segmentation, and fine-grained classification, have used attention mechanism widely and significantly improved. This paper presents an introductory summary of the mechanisms in computer vision problems through recent research. Provide readers with the basics of this widely used method, explore its manifestations under different tasks, evaluate the performance of attention mechanisms, and look forward to its development trends.