Feature Selection Module for CNN Based Object Detector

In the field of computer vision, detection multiple objects with different scales within a single image is challenging. To target this problem, feature pyramids are basic component commonly found in multi-scale object detectors. construction standard pyramids, semantic features simply connected to rebuild new map, regardless whether these have positive effect output or not. order avoid introducing too many redundant fusion stage, module called Feature Selection Module (FSM) was proposed paper, which can automatically detect most representative for rebuilding maps. The channel attention mechanism FSM able process and score each channel, filtering out irrelevant while focusing on high contribution. Moreover, be easily embedded pyramids. Simply adding small number trainable parameters network significantly improve ability extraction. We validated our VOC 2007 dataset, based Yolo series Findings from present study demonstrates that computational cost, method consistently performance