Research on the Algorithm with or without Anchor Frame
Meijun Zheng,Yimin Tian,Shuai Yang,Yunfei Du,Hongmei Chen,Fangfang Song,Xue Gao,Jian Hou
DOI: https://doi.org/10.1109/MLBDBI54094.2021.00068
2021-01-01
Abstract:Target detection is the core and one of the most challenging problems in the field of computer vision research, and it is also a popular direction in computer vision and digital image processing. It is widely used in robots and car navigation, industrial inspection, aerospace, vehicle obstacle avoidance and other scenes. Human consumption can also be reduced by computer vision. With the widespread application of deep learning technology, the efficiency and accuracy of target detection are gradually improved, and it is also for subsequent tasks, such as face recognition, gait recognition, crowd counting, and semantic segmentation. Waiting plays an important cornerstone role. In some respects, it reaches or exceeds the resolution level of the human eye. With the development of computing power and deep learning, the current environment puts forward new requirements for target detection, which improves the accuracy of small target detection. Therefore, this article compares the target detection algorithms with or without anchor frames, analyzes and compares various algorithms, looks forward to the future development trend of the target detection field, and several important directions that can be studied.