90,436 Hits in 6.2 sec

Depth Completion Using a View-constrained Deep Prior [article]

Pallabi Ghosh, Vibhav Vineet, Larry S. Davis, Abhinav Shrivastava, Sudipta Sinha, Neel Joshi
2020 arXiv   pre-print
with a view-constrained photo-consistency loss.  ...  We apply this deep depth prior for inpainting and refining incomplete and noisy depth maps within both binocular and multi-view stereo pipelines.  ...  Using a network similar to Ulyanov et al., our approach generates a depth map by combining a depth reconstruction loss with a view constrained photoconsistency loss.  ... 
arXiv:2001.07791v3 fatcat:23gudl32qbccfnsahnnp6srcby

Scene Intrinsics and Depth from a Single Image

Evan Shelhamer, Jonathan T. Barron, Trevor Darrell
2015 2015 IEEE International Conference on Computer Vision Workshop (ICCVW)  
This combination yields full decompositions by uniting feature learning through deep network regression with physical modeling through statistical priors and random field regularization.  ...  Our full intrinsic pipeline regresses depth by a fully convolutional network then jointly optimizes the intrinsic factorization to recover the input image.  ...  , and prior approaches to deep image factorization.  ... 
doi:10.1109/iccvw.2015.39 dblp:conf/iccvw/ShelhamerBD15 fatcat:mbuyho7rvvaj3h5hmd6qdwypcu

Distill Knowledge from NRSfM for Weakly Supervised 3D Pose Learning [article]

Chaoyang Wang, Chen Kong, Simon Lucey
2019 arXiv   pre-print
Directly using those wrong depth as hard target would negatively impact the student. Instead, we propose a novel loss that ties depth prediction to the cost function used in NRSfM.  ...  No 3D data, multi-view/temporal footage, or object specific prior is required. This alleviates the data bottleneck, which is one of the major concern for supervised methods.  ...  We do not use any 3D labels, multi-view footage, or target specific shape prior.  ... 
arXiv:1908.06377v1 fatcat:jbnz7jeszjcbpc2uj5ykee67ey

Hands Deep in Deep Learning for Hand Pose Estimation [article]

Markus Oberweger, Paul Wohlhart, Vincent Lepetit
2016 arXiv   pre-print
We introduce and evaluate several architectures for Convolutional Neural Networks to predict the 3D joint locations of a hand given a depth map.  ...  We first show that a prior on the 3D pose can be easily introduced and significantly improves the accuracy and reliability of the predictions.  ...  Our proposed architecture Deep-Prior-ORRef, the constrained pose CNN with refinement stage, provides the highest accuracy.  ... 
arXiv:1502.06807v2 fatcat:5jwjajh2fncfrezkku26xbmkyy

Dense Depth Priors for Neural Radiance Fields from Sparse Input Views [article]

Barbara Roessle, Jonathan T. Barron, Ben Mildenhall, Pratul P. Srinivasan, Matthias Nießner
2022 arXiv   pre-print
Our method aims to synthesize novel views of whole rooms from an order of magnitude fewer images. To this end, we leverage dense depth priors in order to constrain the NeRF optimization.  ...  Second, we use depth completion to convert these sparse points into dense depth maps and uncertainty estimates, which are used to guide NeRF optimization.  ...  Acknowledgements This project is funded by a TUM-IAS Rudolf Mößbauer Fellowship, the ERC Starting Grant Scan2CAD (804724), and the German Research Foundation (DFG) Grant Making Machine Learning on Static  ... 
arXiv:2112.03288v2 fatcat:zq5tdwghyfhhfhzrcbckgvygs4

Human Motion Tracking with Less Constraint of Initial Posture from a Single RGB-D Sensor

Chen Liu, Anna Wang, Chunguang Bu, Wenhui Wang, Haijing Sun
2021 Sensors  
To break the self-scanning constraint, we generated a complete mesh only using the front view input to initialize the geometric capture.  ...  In this paper, we propose a human motion capture system combined with human priors and performance capture that only uses a single RGB-D sensor.  ...  Inner Body Alignment In particular, since we only reconstructed a single human object, using a human prior can constrain the unnatural mesh deformation effectively.  ... 
doi:10.3390/s21093029 pmid:33925847 fatcat:wgzlbubwhnczphvpsqc3rc2gzm

Food Volume Estimation Based on Deep Learning View Synthesis from a Single Depth Map

Frank Lo, Yingnan Sun, Jianing Qiu, Benny Lo
2018 Nutrients  
A distinct neural network is designed to use a depth image from one viewing angle to predict another depth image captured from the corresponding opposite viewing angle.  ...  In this paper, a view synthesis approach based on deep learning is proposed to reconstruct 3D point clouds of food items and estimate the volume from a single depth image.  ...  Depth images from two viewing angles (A,B) were used to carry out point cloud completion, respectively.  ... 
doi:10.3390/nu10122005 fatcat:kt3mxsztw5dajoxqzh6lthi2ie

Deep-SLAM++: Object-level RGBD SLAM based on class-specific deep shape priors [article]

Lan Hu, Wanting Xu, Kun Huang, Laurent Kneip
2019 arXiv   pre-print
We propose a discrete selection strategy that finds the best among multiple proposals from different registered views by re-enforcing the agreement with the online depth measurements.  ...  The difficulty of using such predictions in a real SLAM scenario is that the prediction performance depends on the view-point and measurement quality, with even small changes of the input data sometimes  ...  Though Deep-SLAM++ also needs depth measurements to discriminate model candidates, it proves the ability to do this with a very limited amount of correct depth measurements.  ... 
arXiv:1907.09691v2 fatcat:4p6apklvrnc6hm3yin7mxwmska

Dense Depth Posterior (DDP) From Single Image and Sparse Range

Yanchao Yang, Alex Wong, Stefano Soatto
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
We exploit a Conditional Prior Network, that allows associating a probability to each depth value given an image, and combine it with a likelihood term that uses the sparse measurements.  ...  We test our approach on both unsupervised and supervised depth completion using the KITTI benchmark, and improve the state-ofthe-art in both.  ...  Recent works in view synthesis [10, 33] and unsupervised single image depth prediction, [8, 12, 13, 32] propose using view synthesis to hallucinate a novel view image by reconstruction loss.  ... 
doi:10.1109/cvpr.2019.00347 dblp:conf/cvpr/0001WS19 fatcat:vqmkljj3bbbrddly64ozha2wda

Multi-View Silhouette and Depth Decomposition for High Resolution 3D Object Representation [article]

Edward Smith, Scott Fujimoto, David Meger
2018 arXiv   pre-print
Drawing motivation from the canonical view representation of objects, we introduce a novel method for the fast up-sampling of 3D objects in voxel space through networks that perform super-resolution on  ...  We consider the problem of scaling deep generative shape models to high-resolution.  ...  The output of the constrained depth map and silhouette networks are then combined to produce a complete prediction for the high-resolution ODM.  ... 
arXiv:1802.09987v3 fatcat:qj27gfxsxnawrozxhfk77beada

High Accuracy Face Geometry Capture using a Smartphone Video [article]

Shubham Agrawal, Anuj Pahuja, Simon Lucey
2020 arXiv   pre-print
Our method reconstructs accurate face geometry of a subject using a video shot from a smartphone in an unconstrained environment.  ...  By not being constrained to a model subspace, our reconstructed meshes capture important details while being robust to noise and being topologically consistent.  ...  [22] and use 12 source views for each reference view for depth inference. The multiview PatchMatch estimates a depth map for each of the keyframes.  ... 
arXiv:2003.08583v1 fatcat:ttr5s3o77rgwdi24un6ubux42u

Function4D: Real-time Human Volumetric Capture from Very Sparse Consumer RGBD Sensors [article]

Tao Yu, Zerong Zheng, Kaiwen Guo, Pengpeng Liu, Qionghai Dai, Yebin Liu
2021 arXiv   pre-print
Moreover, for detailed and complete surface generation, we propose detail-preserving deep implicit functions for RGBD input which can not only preserve the geometric details on the depth inputs but also  ...  In this paper, we propose a human volumetric capture method that combines temporal volumetric fusion and deep implicit functions.  ...  The goal of the deep implicit surface reconstruction step is to generate complete and detailed surface reconstruction results using deep implicit functions.  ... 
arXiv:2105.01859v2 fatcat:e7atvbmzqrbs3jbhyflkdpscdq

Dense Depth Posterior (DDP) from Single Image and Sparse Range [article]

Yanchao Yang, Alex Wong, Stefano Soatto
2019 arXiv   pre-print
We exploit a Conditional Prior Network, that allows associating a probability to each depth value given an image, and combine it with a likelihood term that uses the sparse measurements.  ...  We test our approach on both unsupervised and supervised depth completion using the KITTI benchmark, and improve the state-of-the-art in both.  ...  Recent works in view synthesis [9, 32] and unsupervised single image depth prediction, [11, 12] propose using view synthesis to hallucinate a novel view image by reconstruction loss.  ... 
arXiv:1901.10034v2 fatcat:eiobbnovyzgtzc5enpmn4pnjou

Deep Octree-based CNNs with Output-Guided Skip Connections for 3D Shape and Scene Completion [article]

Peng-Shuai Wang and Yang Liu and Xin Tong
2020 arXiv   pre-print
We present a simple yet effective deep learning approach for completing the input noisy and incomplete shapes or scenes.  ...  Acquiring complete and clean 3D shape and scene data is challenging due to geometric occlusion and insufficient views during 3D capturing.  ...  Semantic scene completion from a depth image The goal is to predict the occupancy and semantic labels in the view frustum for single depth images of indoor scenes.  ... 
arXiv:2006.03762v1 fatcat:sjakew3nl5d7xmeol4vrskcejy

Modeling Facial Geometry Using Compositional VAEs

Timur Bagautdinov, Chenglei Wu, Jason Saragih, Pascal Fua, Yaser Sheikh
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
We propose a method for learning non-linear face geometry representations using deep generative models.  ...  This parameterization enables us to do model fitting while capturing varying level of detail under different types of geometrical constraints.  ...  a prior on the posterior q.  ... 
doi:10.1109/cvpr.2018.00408 dblp:conf/cvpr/BagautdinovWSFS18 fatcat:n36u3crchfelphgtb3kpxtwo6e
« Previous Showing results 1 — 15 out of 90,436 results