Due to the image multi-target detection and segmentation algorithm is constrained by the low robustness of multi-target detection, and the correlation between the target detection result and the target segmentation… Click to show full abstract
Due to the image multi-target detection and segmentation algorithm is constrained by the low robustness of multi-target detection, and the correlation between the target detection result and the target segmentation method is relatively weak, so accurate and robust multi-target detection and segmentation results cannot be obtained at the same time. In this paper, a multi target detection and segmentation algorithm for fast intelligent network based on regional suggestion is proposed. In order to verify the effectiveness of the algorithm, experiments are carried out on COCO and Cityscapes datasets. The algorithm and the existing methods are compared in the accuracy and time efficiency of IoU. Among them, the mAP of the target detection has been improved by 5.68%, and the target segmentation has been improved by 3.79%. Due to the detection and segmentation of the algorithm is conducted simultaneously in this paper, the time complexity is further improved. The algorithm proposed in this paper can achieve multi-target detection and segmentation simultaneously in the framework of the region’s fast depth model. The accuracy and efficiency have been greatly improved, and it has application prospects in various image processing occasions.
               
Click one of the above tabs to view related content.