Fig. 2
From: Researching on insulator defect recognition based on context cluster CenterNet++
Context Cluster CenterNet++ framework. The flowchart at the top illustrates the overall structure of the proposed network. The input image is first processed by the CCR backbone for feature extraction, followed by feature fusion using PAN, and finally, the detection results are obtained through the Detector. The three subfigures at the bottom represent the overall structures of PAN, Detector, and CCR, from left to right.
