A Multi-view Deep Learning Approach for Detecting Threats on 3D Human Body

摘要

Deep Neural Network-based methods have recently shown an outstanding performance on object detection tasks in 2D scenarios. But many tasks in real world requires object detection in 3D space. In order to narrow this gap, we investigate the task of detection and localization in 3D human body in this paper, and propose a multi-view-based deep learning approach to solve this issue. The experiments show that the proposed approach can effectively detect and locate specific stuff in 3D human body with high accuracy.