{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/aerial-single-view-depth-completion-with","title":"Aerial Single-View Depth Completion with Image-Guided Uncertainty Estimation","arxiv_id":null,"date":"2020-01-17","proceeding":null,"authors":["Lucas Teixeira","Martin R. Oswald","Marc Pollefeys","Margarita Chli"],"abstract":"On the pursuit of autonomous flying robots, the scientific community\r\nhas been developing onboard real-time algorithms for localisation,\r\nmapping and planning. Despite recent progress, the available solutions\r\nstill lack accuracy and robustness in many aspects. While mapping for\r\nautonomous cars had a substantive boost using deep-learning techniques\r\nto enhance LIDAR measurements using image-based depth completion, the\r\nlarge viewpoint variations experienced by aerial vehicles are still\r\nposing major challenges for learning-based mapping approaches. In this\r\npaper, we propose a depth completion and uncertainty estimation\r\napproach that better handles the challenges of aerial platforms, such\r\nas large viewpoint and depth variations, and limited computing\r\nresources. The core of our method is a novel compact network that\r\nperforms both depth completion and confidence estimation using an\r\nimage-guided approach. Real-time performance onboard a GPU suitable for\r\nsmall flying robots is achieved by sharing deep features between both\r\ntasks. Experiments demonstrate that our network outperforms the\r\nstate-of-the-art in depth completion and uncertainty estimation for\r\nsingle-view methods on mobile GPUs. We further present a new\r\nphotorealistic aerial depth completion dataset that exhibits more\r\nchallenging depth completion scenarios than the established indoor and\r\ncar driving datasets. The dataset includes an open-source,\r\nvisual-inertial UAV simulator for photo-realistic data generation. Our\r\nresults show that our network trained on this dataset can be directly\r\ndeployed on real-world outdoor aerial public datasets without\r\nfine-tuning or style transfer.","url_abs":"https://doi.org/10.3929/ethz-b-000392181","url_pdf":"https://www.research-collection.ethz.ch/bitstream/handle/20.500.11850/392181/19-1175_03_MS.pdf?sequence=1&isAllowed=y","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"aerial-single-view-depth-completion-with","repo_url":"https://github.com/VIS4ROB-lab/aerial-depth-completion","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"aerial-single-view-depth-completion-with","repo_url":"https://github.com/VIS4ROB-lab/visensor_simulator","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"depth-completion","task_name":"Depth Completion"},{"task_slug":null,"task_name":"GPU"},{"task_slug":"monocular-depth-estimation","task_name":"Monocular Depth Estimation"},{"task_slug":"style-transfer","task_name":"Style Transfer"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}