Filters








25,077 Hits in 4.4 sec

Improving Optical Flow on a Pyramid Level [article]

Markus Hofinger, Samuel Rota Bulò, Lorenzo Porzi, Arno Knapitsch, Thomas Pock, Peter Kontschieder
2020 arXiv   pre-print
Within an individual pyramid level, we improve the cost volume construction process by departing from a warping- to a sampling-based strategy, which avoids ghosting and hence enables us to better preserve  ...  Our second contribution revises the gradient flow across pyramid levels. The typical operations performed at each pyramid level can lead to noisy, or even contradicting gradients across levels.  ...  Improving Optical Flow on a Pyramid Level - Supplementary Material This document contains supplementary material for the paper 'Improving Op- tical Flow on a Pyramid Level'.  ... 
arXiv:1912.10739v2 fatcat:2kh62y3uuzcytnjwnh6cd6yn6q

Detail Preserving Residual Feature Pyramid Modules for Optical Flow [article]

Libo Long, Jochen Lang
2021 arXiv   pre-print
Feature pyramids and iterative refinement have recently led to great progress in optical flow estimation.  ...  Results show that our RFPM visibly reduces flow errors and improves state-of-art performance in the clean pass of Sintel, and is one of the top-performing methods in KITTI.  ...  We decide to focus on the feature pyramid because we do not want to change the iterative refinement architecture as it has been shown to improve optical flow results, especially for large flow, and because  ... 
arXiv:2107.10990v1 fatcat:2dnyjzmqazazjosc6n5oymy4c4

UPFlow: Upsampling Pyramid for Unsupervised Optical Flow Learning [article]

Kunming Luo, Chuan Wang, Shuaicheng Liu, Haoqiang Fan, Jue Wang, Jian Sun
2021 arXiv   pre-print
We present an unsupervised learning approach for optical flow estimation by improving the upsampling and learning of pyramid network.  ...  Moreover, we propose a pyramid distillation loss to add supervision for intermediate levels via distilling the finest flow as pseudo labels.  ...  For the top-down guidance of the pyramid network, we proposed a pyramid distillation loss to improve the optical flow learning on intermediate levels of the network.  ... 
arXiv:2012.00212v2 fatcat:eikcy4jnerff5cn3zzdbw64vcu

Optical Flow Estimation using a Spatial Pyramid Network [article]

Anurag Ranjan, Michael J. Black
2016 arXiv   pre-print
We learn to compute optical flow by combining a classical spatial-pyramid formulation with deep learning.  ...  This estimates large motions in a coarse-to-fine approach by warping one image of a pair at each pyramid level by the current flow estimate and computing an update to the flow.  ...  Acknowledgement We thank Jonas Wulff for his insightful discussions about optical flow.  ... 
arXiv:1611.00850v2 fatcat:mkel6t3zjnd7dcwvzhbdbd2ypm

Optical Flow Estimation Using a Spatial Pyramid Network

Anurag Ranjan, Michael J. Black
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
We learn to compute optical flow by combining a classical spatial-pyramid formulation with deep learning.  ...  This estimates large motions in a coarse-to-fine approach by warping one image of a pair at each pyramid level by the current flow estimate and computing an update to the flow.  ...  We thank Jonas Wulff for his insightful discussions about optical flow.  ... 
doi:10.1109/cvpr.2017.291 dblp:conf/cvpr/RanjanB17 fatcat:dyjoerjg4bbmdmcva5utwlqob4

FastFlowNet: A Lightweight Network for Fast Optical Flow Estimation [article]

Lingtong Kong, Chunhua Shen, Jie Yang
2021 arXiv   pre-print
Third, an efficient shuffle block decoder (SBD) is implanted into each pyramid level to accelerate flow estimation with marginal drops in accuracy.  ...  Dense optical flow estimation plays a key role in many robotic vision tasks. In the past few years, with the advent of deep learning, we have witnessed great progress in optical flow estimation.  ...  Most recent works on optical flow focus on improving accuracy. IRR-PWC [15] shares the flow decoder and context network among all spatial scales for joint optical flow and occlusion estimation.  ... 
arXiv:2103.04524v2 fatcat:rpedqrc3f5bjjpyeljno6jywre

Models Matter, So Does Training: An Empirical Study of CNNs for Optical Flow Estimation [article]

Deqing Sun, Xiaodong Yang, Ming-Yu Liu, Jan Kautz
2018 arXiv   pre-print
We investigate two crucial and closely related aspects of CNNs for optical flow estimation: models and training.  ...  We further improve the training procedure and increase the accuracy of PWC-Net on Sintel by 10\% and on KITTI 2012 and 2015 by 20\%.  ...  Warping allows for estimating a small optical flow (increment) at each pyramid level to deal with a large optical flow.  ... 
arXiv:1809.05571v1 fatcat:hzpvkwnxjnehjp2m6zwq7tye4m

FDFlowNet: Fast Optical Flow Estimation using a Deep Lightweight Network [article]

Lingtong Kong, Jie Yang
2020 arXiv   pre-print
In this work, we present a lightweight yet effective model for real-time optical flow estimation, termed FDFlowNet (fast deep flownet).  ...  We first introduce an U-shape network for constructing multi-scale feature which benefits upper levels with global receptive field compared with pyramid network.  ...  In summary our contributions are: • We propose a compact and effecient U-shape network as a improvement of pyramid network for optical flow estimation that can efficiently fuse multi-scale information  ... 
arXiv:2006.12263v1 fatcat:g5kt4xusj5eaxkpdbcovxtdck4

A Lightweight Optical Flow CNN - Revisiting Data Fidelity and Regularization [article]

Tak-Wai Hui, Xiaoou Tang, Chen Change Loy
2020 arXiv   pre-print
We compute optical flow in a spatial-pyramid formulation as SPyNet but through a novel lightweight cascaded flow inference.  ...  Comparing to LiteFlowNet, LiteFlowNet2 improves the optical flow accuracy on Sintel Clean by 23.3%, Sintel Final by 12.8%, KITTI 2012 by 19.6%, and KITTI 2015 by 18.8%, while being 2.2 times faster.  ...  feature encoder, NetE: a multi-scale flow decoder) and pyramid levels of LiteFlowNet [13] trained on Things3D.  ... 
arXiv:1903.07414v3 fatcat:xoftivn44rdsdch7hzzrtg52ou

PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume [article]

Deqing Sun, Xiaodong Yang, Ming-Yu Liu, Jan Kautz
2018 arXiv   pre-print
Cast in a learnable feature pyramid, PWC-Net uses the cur- rent optical flow estimate to warp the CNN features of the second image.  ...  We present a compact but effective CNN model for optical flow, called PWC-Net.  ...  Warping allows for estimating a small optical flow (increment) at each pyramid level to deal with a large optical flow.  ... 
arXiv:1709.02371v3 fatcat:qmxjgwd65bdsxikpnsifiguu7a

Accurate Realtime Motion Estimation Using Optical Flow on an Embedded System

Anis Ammar, Hana Ben Fredj, Chokri Souani
2021 Electronics  
Our approach was applied on a local treatment region implemented into Raspberry Pi 4, with several improvements.  ...  In this context, we designed an accurate motion estimation system based on the calculation of the optical flow of a moving object using the Lucas–Kanade algorithm.  ...  flow refinement computing final optical flow output: final optical flow From two successive images, we created a Gaussian pyramid for each image.  ... 
doi:10.3390/electronics10172164 fatcat:26o6f6i7nzfyzkplqvsku2x2oq

PWOC-3D: Deep Occlusion-Aware End-to-End Scene Flow Estimation [article]

Rohan Saxena, René Schuster, Oliver Wasenmüller, Didier Stricker
2019 arXiv   pre-print
In the last few years, convolutional neural networks (CNNs) have demonstrated increasing success at learning many computer vision tasks including dense estimation problems such as optical flow and stereo  ...  The work presented in this paper overcomes these drawbacks efficiently (in terms of speed and accuracy) by proposing PWOC-3D, a compact CNN architecture to predict scene flow from stereo image sequences  ...  The incorporation of a spatial pyramid and warping at different pyramid levels in an end-to-end CNN for optical flow estimation was first introduced in SPyNet [26] .  ... 
arXiv:1904.06116v1 fatcat:rsvgzb3erbe23mcvt4ekus3moe

FPCR-Net: Feature Pyramidal Correlation and Residual Reconstruction for Optical Flow Estimation [article]

Xiaolin Song, Yuyang Zhao, Jingyu Yang, Cuiling Lan, Wenjun Zeng
2021 arXiv   pre-print
To exploit such flexible and comprehensive information, we propose a semi-supervised Feature Pyramidal Correlation and Residual Reconstruction Network (FPCR-Net) for optical flow estimation from frame  ...  Based on the pyramid correlation mapping, we further propose a correlation-warping-normalization (CWN) module to efficiently exploit the correlation dependency.  ...  We leverage a pyramid correlation mapping operation based on single correlation for aggregating different level cost volumes.  ... 
arXiv:2001.06171v4 fatcat:xrjrstzpmfc5dfdpjgxvlvygny

Optimal Filter Estimation for Lucas-Kanade Optical Flow

Nusrat Sharmin, Remus Brad
2012 Sensors  
Generally, in optical flow computation, filtering is used at the initial level on original input images and afterwards, the images are resized.  ...  In this paper, we propose an image filtering approach as a pre-processing step for the Lucas-Kanade pyramidal optical flow algorithm.  ...  and all levels on pyramidal Lucas-Kanade optical flow.  ... 
doi:10.3390/s120912694 fatcat:qsjdkrsljrbovdf3f4bxnykz6a

PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume

Deqing Sun, Xiaodong Yang, Ming-Yu Liu, Jan Kautz
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
Cast in a learnable feature pyramid, PWC-Net uses the current optical flow estimate to warp the CNN features of the second image.  ...  We present a compact but effective CNN model for optical flow, called PWC-Net.  ...  Warping allows for estimating a small optical flow (increment) at each pyramid level to deal with a large optical flow.  ... 
doi:10.1109/cvpr.2018.00931 dblp:conf/cvpr/SunY0K18 fatcat:spjcu4fnvzepfiwqm4474cy5be
« Previous Showing results 1 — 15 out of 25,077 results