Accelerated DNA-SLAM for RGB-D images

Accelerated DNA-SLAM for RGB-D images
Mina Ameli, Oliver Wasenmüller, Mohammad Reza Soheili, Jamshid Shanbehzadeh, Didier Stricker
ACM International Conference on Image and Graphics Processing (ICIGP-2018), February 24-26, Hong Kong, China

Abstract:
In the highly active research field of Simultaneous Localization And Mapping (SLAM), RGB-D images have been a major interest to use. Real-time SLAM for RGB-D images is of great importance since dense methods using all the depth and intensity values showed superior performance in the past. Due to development of GPU and CPU technologies, the real-time implementation of the mentioned algorithms is no longer an impassable problem. In this paper, we present an acceleration approach for the DNA-SLAM algorithm. We argue some possible challenges while converting the CPU implemented algorithm to the GPU. Finally, runtime evaluation and improvements are shown on the public CoRBS dataset.