Video Object Detection Using Event-Aware Convolutional Lstm and Object Relation Networks

Common video-based object detectors exploit temporal contextual information to improve the performance of detection. However, detecting objects under challenging conditions has not been thoroughly studied yet. In this paper, we focus on improving detection for events such as aspect ratio change, occlusion, or large motion. To end, propose a video network using event-aware ConvLSTM and relation networks. Our proposed is able highlight area where those take place. Compared with traditional ConvLSTM, method it easier support events. further performance, an module supporting frame selection applied enhance pooled features target ROI. It effectively selects same from one reference frames rather than all them. Experimental results ImageNet VID dataset show that achieves mAP 81.0% without any post processing can handle efficiently in