2020/08/04 by Hyungtae Lim, Lim, Hyungtae, Hyeonjae Gil +3 · 1 citation
Computer Science · Earth and Planetary Sciences · #3D Surveying and Cultural Heritage #Advanced Vision and Imaging #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #I.2.9 #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Optical measurement and interference techniques #Robotics (cs.RO) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2008.01405
openalex publication_date 2020/08/04 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
In this study, a deep-learning-based multi-stage network architecture called\nMulti-Stage Depth Prediction Network (MSDPN) is proposed to predict a dense\ndepth map using a 2D LiDAR and a monocular camera. Our proposed network\nconsists of a multi-stage encoder-decoder architecture and Cross Stage Feature\nAggregation (CSFA). The proposed multi-stage encoder-decoder architecture\nalleviates the partial observation problem caused by the characteristics of a\n2D LiDAR, and CSFA prevents the multi-stage network from diluting the features\nand allows the network to learn the inter-spatial relationship between features\nbetter. Previous works use sub-sampled data from the ground truth as an input\nrather than actual 2D LiDAR data. In contrast, our approach trains the model\nand conducts experiments with a physically-collected 2D LiDAR dataset. To this\nend, we acquired our own dataset called KAIST RGBD-scan dataset and validated\nthe effectiveness and the robustness of MSDPN under realistic conditions. As\nverified experimentally, our network yields promising performance against\nstate-of-the-art methods. Additionally, we analyzed the performance of\ndifferent input methods and confirmed that the reference depth map is robust in\nuntrained scenarios.\n