2021/03/31 by Christoph Angermann, Angermann, Christoph, Adéla Moravová +7
Computer Science · #Advanced Image Processing Techniques #Advanced Vision and Imaging #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Optical measurement and interference techniques #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2103.16938
openalex publication_date 2021/03/31 · openalex created_date 2022/07/13 · openalex updated_date 2026/07/28
Real-time estimation of actual environment depth is an essential module for\nvarious autonomous system tasks such as localization, obstacle detection and\npose estimation. During the last decade of machine learning, extensive\ndeployment of deep learning methods to computer vision tasks yielded successful\napproaches for realistic depth synthesis out of a simple RGB modality. While\nmost of these models rest on paired depth data or availability of video\nsequences and stereo images, there is a lack of methods facing single-image\ndepth synthesis in an unsupervised manner. Therefore, in this study, latest\nadvancements in the field of generative neural networks are leveraged to fully\nunsupervised single-image depth synthesis. To be more exact, two\ncycle-consistent generators for RGB-to-depth and depth-to-RGB transfer are\nimplemented and simultaneously optimized using the Wasserstein-1 distance. To\nensure plausibility of the proposed method, we apply the models to a self\nacquised industrial data set as well as to the renown NYU Depth v2 data set,\nwhich allows comparison with existing approaches. The observed success in this\nstudy suggests high potential for unpaired single-image depth estimation in\nreal world applications.\n