Filters








248 Hits in 7.7 sec

Superpixel-based Two-view Deterministic Fitting for Multiple-structure Data [article]

Guobao Xiao, Hanzi Wang, Yan Yan, David Suter
2016 arXiv   pre-print
This paper proposes a two-view deterministic geometric model fitting method, termed Superpixel-based Deterministic Fitting (SDF), for multiple-structure data.  ...  Experimental results demonstrate that the proposed SDF shows superiority over several state-of-the-art fitting methods for real images with single-structure and multiple-structure data.  ...  Then, based on the grouping cues, we propose an efficient and effective method, called Superpixel-based Deterministic Fitting (SDF), for multiple-structure data.  ... 
arXiv:1607.05839v1 fatcat:dtzyswny5newdch2b3md7xc7l4

Superpixel-guided Two-view Deterministic Geometric Model Fitting [article]

Guobao Xiao, Hanzi Wang, Yan Yan, David Suter
2018 arXiv   pre-print
In this paper, we propose a novel superpixel-guided two-view geometric model fitting method (called SDF), which can obtain reliable and consistent results for real images.  ...  Geometric model fitting is a fundamental research topic in computer vision and it aims to fit and segment multiple-structure data.  ...  Natural Science Foundation of China under Grants U1605252, 61702431, 61472334 and 61571379, by the China Postdoctoral Science Foundation 2017M620272 and by the Fujian Province Education-Science Project for  ... 
arXiv:1805.01158v1 fatcat:vplcy3e73bgyhl64w2qp3sq2f4

Binary Graph Descriptor for Robust Relocalization on Heterogeneous Data

Xi Wang, Marc Christie, Eric Marchand
2022 IEEE Robotics and Automation Letters  
Generic Multiple Layer Descriptor Applying the deterministic graph embedding has another advantage: each binary digit's order is determined and corresponds to its combination order given a graph structure  ...  Multiple layers of heterogeneous image are structured in a graph representation spanning over neighboring regions of the superpixel, and embedding in a BiG descriptor through graph embedding and binarization  ... 
doi:10.1109/lra.2022.3142854 fatcat:zlcsyppqjvdfxbhvn3zyg7ghdi

Image Segmentation Based on Graph-Cut Models and Probabilistic Graphical Models: A Comparative Study [chapter]

Maedeh Beheshti, Alan Wee-Chung Liew
2014 Communications in Computer and Information Science  
The purpose of this paper is to introduce two recently proposed graph based segmentation methods, namely, graph-cut models (deterministic) and unified graphical model (probabilistic).  ...  Image Segmentation has been one of the most important unsolved problems in computer vision for many years. Recently, there have been great efforts in producing better segmentation algorithms.  ...  Graph based segmentation algorithms can be broadly divided into two types 1-deterministic 2-probabilistic [1] .  ... 
doi:10.1007/978-3-662-45652-1_37 fatcat:k4plovp3mbhxbfycy77fvm44te

Small Sample Learning of Superpixel Classifiers for EM Segmentation [chapter]

Toufiq Parag, Stephen Plaza, Louis Scheffer
2014 Lecture Notes in Computer Science  
In this paper, we propose an interactive learning scheme for the superpixel classifier for EM segmentation.  ...  Pixel and superpixel classifiers have become essential tools for EM segmentation algorithms.  ...  The authors are grateful to Stuart Berg of Janelia Farm Research for his support in software development.  ... 
doi:10.1007/978-3-319-10404-1_49 fatcat:xwykciuwq5gohmjyt4v2itljyi

Interpretable Structure-Evolving LSTM [article]

Xiaodan Liang and Liang Lin and Xiaohui Shen and Jiashi Feng and Shuicheng Yan and Eric P. Xing
2017 arXiv   pre-print
This paper develops a general framework for learning interpretable data representation via Long Short-Term Memory (LSTM) recurrent neural networks over hierarchal graph structures.  ...  Instead of learning LSTM models over the pre-fixed structures, we propose to further learn the intermediate interpretable multi-level graph structures in a progressive and stochastic way from data during  ...  The network can thus learn representations to better fit the hidden structure of the data.  ... 
arXiv:1703.03055v1 fatcat:qlb32jzyyfgbnnhcynxcyjodgi

Cotemporal Multi-View Video Segmentation

Abdelaziz Djelouah, Jean-Sebastien Franco, Edmond Boyer, Patrick Perez, George Drettakis
2016 2016 Fourth International Conference on 3D Vision (3DV)  
Multi-view methods for dynamic scenes usually rely on geometric calibration to impose spatial shape constraints between viewpoints.  ...  We also take advantage of learning based segmentation strategies by casting the problem as the selection of monocular proposals that satisfy multi-view constraints.  ...  A graph based approach for multi-view region matching that exploits motion descriptors. 2.  ... 
doi:10.1109/3dv.2016.45 dblp:conf/3dim/DjelouahFBPD16 fatcat:3ktqi6kchvgbdbwkhvdlq6lw2m

Small Sample Learning of Superpixel Classifiers for EM Segmentation- Extended Version [article]

Toufiq Parag, Stephen Plaza, Louis Scheffer (Janelia Farm Research Campus- HHMI)
2014 arXiv   pre-print
In this paper, we propose an interactive learning scheme for the superpixel classifier for EM segmentation.  ...  Pixel and superpixel classifiers have become essential tools for EM segmentation algorithms.  ...  Acknowledgement: The authors are grateful to Stuart Berg of Janelia Farm Research for his support in software development.  ... 
arXiv:1406.1774v2 fatcat:ukdxuwtcsvh2dox3ftes2gkvlu

Interpretable Structure-Evolving LSTM

Xiaodan Liang, Liang Lin, Xiaohui Shen, Jiashi Feng, Shuicheng Yan, Eric P. Xing
2017 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
This paper develops a general framework for learning interpretable data representation via Long Short-Term Memory (LSTM) recurrent neural networks over hierarchal graph structures.  ...  Instead of learning LSTM models over the pre-fixed structures, we propose to further learn the intermediate interpretable multi-level graph structures in a progressive and stochastic way from data during  ...  To learn intermediate interpretable graph structures of the data and alleviate the over-fitting problem, Figure 1 .  ... 
doi:10.1109/cvpr.2017.234 dblp:conf/cvpr/LiangLSFYX17 fatcat:bemkznbv25fynkddzh6h4vwphy

Single-View 3D Scene Parsing by Attributed Grammar

Xiaobai Liu, Yibiao Zhao, Song-chun Zhu
2014 2014 IEEE Conference on Computer Vision and Pattern Recognition  
The grammar takes superpixels as its terminal nodes and use five production rules to generate the scene into a hierarchical parse graph.  ...  We also demonstrate that our method is able to recover detailed 3D model with relaxed Manhattan structures which clearly advances the state-of-the-arts of singleview 3D reconstruction.  ...  For the edge e =< a, b >, φ s e is measured using following steps: i) estimate the boundary pixels between two superpixels a and b; ii) fit one straight line, denoted as ab, from the boundary pixels by  ... 
doi:10.1109/cvpr.2014.93 dblp:conf/cvpr/LiuZZ14 fatcat:maumg7rftje3hlvat6bdftc6ha

Fast Interactive Video Object Segmentation with Graph Neural Networks [article]

Viktor Varga, András Lőrincz
2021 arXiv   pre-print
In this paper we present a graph neural network based approach for tackling the problem of interactive video object segmentation.  ...  Pixelwise annotation of image sequences can be very tedious for humans.  ...  For example, one may consider the refinement of superpixel segmentation based on the GNN label predictions after each interaction round.  ... 
arXiv:2103.03821v2 fatcat:ell6h6o6rfhgdccin7wrja4my4

The Random Cluster Model for robust geometric fitting

Trung Thanh Pham, Tat-Jun Chin, Jin Yu, D. Suter
2012 2012 IEEE Conference on Computer Vision and Pattern Recognition  
To take advantage of our hypothesis generator, we construct a simple annealing method based on graph cuts to fit multiple instances of the geometric model in the data.  ...  Experimental results show clear improvements in efficiency over other methods based on minimal subset samplers.  ...  ( Two-view motion segmentation. Given two views of a dynamic scene, a number of feature points are extracted and matched across two images.  ... 
doi:10.1109/cvpr.2012.6247740 dblp:conf/cvpr/PhamCYS12 fatcat:iai2mtv3kzemxevqubky6wolhe

The Random Cluster Model for Robust Geometric Fitting

Trung T. Pham, Tat-Jun Chin, Jin Yu, David Suter
2014 IEEE Transactions on Pattern Analysis and Machine Intelligence  
To take advantage of our hypothesis generator, we construct a simple annealing method based on graph cuts to fit multiple instances of the geometric model in the data.  ...  Experimental results show clear improvements in efficiency over other methods based on minimal subset samplers.  ...  ( Two-view motion segmentation. Given two views of a dynamic scene, a number of feature points are extracted and matched across two images.  ... 
doi:10.1109/tpami.2013.2296310 pmid:26353345 fatcat:hsdmof2jlzestjlk5rtxbi7lji

GEOMETRIC CONTEXT AND ORIENTATION MAP COMBINATION FOR INDOOR CORRIDOR MODELING USING A SINGLE IMAGE

Ali Baligh Jahromi, Gunho Sohn
2016 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences  
Considering the available data gathering techniques with respect to the sensors cost and data processing time, single images proved to be one of the reliable sources.  ...  Many of the current single image based indoor space modeling methods are defining the scene as a single box primitive.  ...  The common solution is to build structural knowledge of the image from pixels to superpixels. Hoiem et al., (2007) solution was to compute multiple segmentations based on simple cues.  ... 
doi:10.5194/isprs-archives-xli-b4-295-2016 fatcat:v4qggqa645b67krvb6kj6ipfzi

Beyond pixels: A comprehensive survey from bottom-up to semantic image segmentation and cosegmentation

Hongyuan Zhu, Fanman Meng, Jianfei Cai, Shijian Lu
2016 Journal of Visual Communication and Image Representation  
Finally, we suggest some design flavors and research directions for future research in image segmentation.  ...  Covering around 180 publications, we give an overview of broad areas of segmentation topics including not only the classic bottom-up approaches, but also the recent development in superpixel, interactive  ...  superpixels at structure-less regions.  ... 
doi:10.1016/j.jvcir.2015.10.012 fatcat:siynj55yxrdqlj5vcrpfm7nyhe
« Previous Showing results 1 — 15 out of 248 results