Group-Based Recursive Learning for Multistage Object Detection

Object detection deals by detecting multiple instances of semantic objects of a specific class in digital photographs. Bounding box is one of them, used to describe the spatial location of an object. Among proposals, the significant and essential spatial layout correlations are largely neglected but are still crucial for precise object recognition. The proposed method repeatedly enhances object detection using a multistage architecture and group recursive learning technique for decision making process. The two techniques—weak supervised object segmentation and recursive detection refinement—are applied in this study. The above combination yields effective results in object detection on PASCAL VOC-2K7 and VOC-2K12 datasets.