Matches in SemOpenAlex for { <https://semopenalex.org/work/W2896853695> ?p ?o ?g. }
- W2896853695 abstract "Learning to estimate 3D geometry in a single frame and optical flow from consecutive frames by watching unlabeled videos via deep convolutional network has made significant progress recently. Current state-of-the-art (SoTA) methods treat the two tasks independently. One typical assumption of the existing depth estimation methods is that the scenes contain no independent moving objects. while object moving could be easily modeled using optical flow. In this paper, we propose to address the two tasks as a whole, i.e. to jointly understand per-pixel 3D geometry and motion. This eliminates the need of static scene assumption and enforces the inherent geometrical consistency during the learning process, yielding significantly improved results for both tasks. We call our method as Every Pixel Counts++ or EPC++. Specifically, during training, given two consecutive frames from a video, we adopt three parallel networks to predict the camera motion (MotionNet), dense depth map (DepthNet), and per-pixel optical flow between two frames (OptFlowNet) respectively. The three types of information are fed into a holistic 3D motion parser (HMP), and per-pixel 3D motion of both rigid background and moving objects are disentangled and recovered. Comprehensive experiments were conducted on datasets with different scenes, including driving scenario (KITTI 2012 and KITTI 2015 datasets), mixed outdoor/indoor scenes (Make3D) and synthetic animation (MPI Sintel dataset). Performance on the five tasks of depth estimation, optical flow estimation, odometry, moving object segmentation and scene flow estimation shows that our approach outperforms other SoTA methods. Code will be available at: https://github.com/chenxuluo/EPC." @default.
- W2896853695 created "2018-10-26" @default.
- W2896853695 creator A5003629926 @default.
- W2896853695 creator A5016423364 @default.
- W2896853695 creator A5029181844 @default.
- W2896853695 creator A5049078993 @default.
- W2896853695 creator A5080385311 @default.
- W2896853695 creator A5083556808 @default.
- W2896853695 creator A5086706224 @default.
- W2896853695 date "2018-10-14" @default.
- W2896853695 modified "2023-10-16" @default.
- W2896853695 title "Every Pixel Counts ++: Joint Learning of Geometry and Motion with 3D Holistic Understanding" @default.
- W2896853695 cites W1496571393 @default.
- W2896853695 cites W1513100184 @default.
- W2896853695 cites W1686810756 @default.
- W2896853695 cites W1836465849 @default.
- W2896853695 cites W1899309388 @default.
- W2896853695 cites W1903029394 @default.
- W2896853695 cites W1905829557 @default.
- W2896853695 cites W1907877624 @default.
- W2896853695 cites W1915250530 @default.
- W2896853695 cites W1921093919 @default.
- W2896853695 cites W1951289974 @default.
- W2896853695 cites W1964057156 @default.
- W2896853695 cites W1992178727 @default.
- W2896853695 cites W1994804971 @default.
- W2896853695 cites W2074254947 @default.
- W2896853695 cites W2091630150 @default.
- W2896853695 cites W2100315781 @default.
- W2896853695 cites W2108134361 @default.
- W2896853695 cites W2110483288 @default.
- W2896853695 cites W2113107168 @default.
- W2896853695 cites W2121299550 @default.
- W2896853695 cites W2124907686 @default.
- W2896853695 cites W2125310925 @default.
- W2896853695 cites W2125416623 @default.
- W2896853695 cites W2129671742 @default.
- W2896853695 cites W2132947399 @default.
- W2896853695 cites W2133665775 @default.
- W2896853695 cites W2150066425 @default.
- W2896853695 cites W2150839555 @default.
- W2896853695 cites W2158211626 @default.
- W2896853695 cites W2168538937 @default.
- W2896853695 cites W2171740948 @default.
- W2896853695 cites W2216090464 @default.
- W2896853695 cites W2300779272 @default.
- W2896853695 cites W2336968928 @default.
- W2896853695 cites W2402395722 @default.
- W2896853695 cites W2460260369 @default.
- W2896853695 cites W2493152400 @default.
- W2896853695 cites W2507953016 @default.
- W2896853695 cites W2520707372 @default.
- W2896853695 cites W253426031 @default.
- W2896853695 cites W2548527721 @default.
- W2896853695 cites W2550402137 @default.
- W2896853695 cites W2560474170 @default.
- W2896853695 cites W2585592883 @default.
- W2896853695 cites W2604909019 @default.
- W2896853695 cites W2608018946 @default.
- W2896853695 cites W2609883120 @default.
- W2896853695 cites W2751023760 @default.
- W2896853695 cites W2779333428 @default.
- W2896853695 cites W2788857104 @default.
- W2896853695 cites W2810554001 @default.
- W2896853695 cites W2885093229 @default.
- W2896853695 cites W2890949887 @default.
- W2896853695 cites W2894983388 @default.
- W2896853695 cites W2949023359 @default.
- W2896853695 cites W2951333975 @default.
- W2896853695 cites W2962807621 @default.
- W2896853695 cites W2962864875 @default.
- W2896853695 cites W2963128679 @default.
- W2896853695 cites W2963131444 @default.
- W2896853695 cites W2963150697 @default.
- W2896853695 cites W2963170338 @default.
- W2896853695 cites W2963395775 @default.
- W2896853695 cites W2963412495 @default.
- W2896853695 cites W2963549785 @default.
- W2896853695 cites W2963583471 @default.
- W2896853695 cites W2963591054 @default.
- W2896853695 cites W2963654727 @default.
- W2896853695 cites W2963782415 @default.
- W2896853695 cites W2963823258 @default.
- W2896853695 cites W2963891416 @default.
- W2896853695 cites W2963906250 @default.
- W2896853695 cites W2963924401 @default.
- W2896853695 cites W2964121744 @default.
- W2896853695 cites W2964314455 @default.
- W2896853695 cites W2964968086 @default.
- W2896853695 cites W2985775862 @default.
- W2896853695 cites W3199643754 @default.
- W2896853695 cites W337610345 @default.
- W2896853695 cites W612478963 @default.
- W2896853695 cites W764651262 @default.
- W2896853695 doi "https://doi.org/10.48550/arxiv.1810.06125" @default.
- W2896853695 hasPublicationYear "2018" @default.
- W2896853695 type Work @default.
- W2896853695 sameAs 2896853695 @default.
- W2896853695 citedByCount "9" @default.
- W2896853695 countsByYear W28968536952019 @default.