This paper presents a semi-automatic segmentation technique called Bayesian cut that formulates object boundary detection as the most probable explanation (MPE) of a Bayesian network’s joint probability distribution. A two-layer Bayesian network structure is formulated from a planar graph representing a watershed segmentation of an image. The network’s prior probabilities encode the confidence that an edge in the planar graph belongs to an object boundary while the conditional probability tables (CPTs) enforce global contour properties of closure and simplicity (i.e., no self-intersections). Evidence, in the form of one or more connected boundary points, allows the network to compute the MPE with minimal user guidance. The constraints imposed by CPTs also permit a linear-time algorithm to compute the MPE, which in turn allows for interactive segmentation where every mouse movement recomputes the MPE based on the current cursor position and displays the corresponding segmentation.
{"title":"Real-Time Semi-Automatic Segmentation Using a Bayesian Network","authors":"Eric N. Mortensen, J. Jia","doi":"10.1109/CVPR.2006.239","DOIUrl":"https://doi.org/10.1109/CVPR.2006.239","url":null,"abstract":"This paper presents a semi-automatic segmentation technique called Bayesian cut that formulates object boundary detection as the most probable explanation (MPE) of a Bayesian network’s joint probability distribution. A two-layer Bayesian network structure is formulated from a planar graph representing a watershed segmentation of an image. The network’s prior probabilities encode the confidence that an edge in the planar graph belongs to an object boundary while the conditional probability tables (CPTs) enforce global contour properties of closure and simplicity (i.e., no self-intersections). Evidence, in the form of one or more connected boundary points, allows the network to compute the MPE with minimal user guidance. The constraints imposed by CPTs also permit a linear-time algorithm to compute the MPE, which in turn allows for interactive segmentation where every mouse movement recomputes the MPE based on the current cursor position and displays the corresponding segmentation.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131359161","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Part-based tree-structured models have been widely used for 2D articulated human pose-estimation. These approaches admit efficient inference algorithms while capturing the important kinematic constraints of the human body as a graphical model. These methods often fail however when multiple body parts fit the same image region resulting in global pose estimates that poorly explain the overall image evidence. Attempts to solve this problem have focused on the use of strong prior models that are limited to learned activities such as walking. We argue that the problem actually lies with the image observations and not with the prior. In particular, image evidence for each body part is estimated independently of other parts without regard to self-occlusion. To address this we introduce occlusion-sensitive local likelihoods that approximate the global image likelihood using per-pixel hidden binary variables that encode the occlusion relationships between parts. This occlusion reasoning introduces interactions between non-adjacent body parts creating loops in the underlying graphical model. We deal with this using an extension of an approximate belief propagation algorithm (PAMPAS). The algorithm recovers the real-valued 2D pose of the body in the presence of occlusions, does not require strong priors over body pose and does a quantitatively better job of explaining image evidence than previous methods.
{"title":"Measure Locally, Reason Globally: Occlusion-sensitive Articulated Pose Estimation","authors":"L. Sigal, Michael J. Black","doi":"10.1109/CVPR.2006.180","DOIUrl":"https://doi.org/10.1109/CVPR.2006.180","url":null,"abstract":"Part-based tree-structured models have been widely used for 2D articulated human pose-estimation. These approaches admit efficient inference algorithms while capturing the important kinematic constraints of the human body as a graphical model. These methods often fail however when multiple body parts fit the same image region resulting in global pose estimates that poorly explain the overall image evidence. Attempts to solve this problem have focused on the use of strong prior models that are limited to learned activities such as walking. We argue that the problem actually lies with the image observations and not with the prior. In particular, image evidence for each body part is estimated independently of other parts without regard to self-occlusion. To address this we introduce occlusion-sensitive local likelihoods that approximate the global image likelihood using per-pixel hidden binary variables that encode the occlusion relationships between parts. This occlusion reasoning introduces interactions between non-adjacent body parts creating loops in the underlying graphical model. We deal with this using an extension of an approximate belief propagation algorithm (PAMPAS). The algorithm recovers the real-valued 2D pose of the body in the presence of occlusions, does not require strong priors over body pose and does a quantitatively better job of explaining image evidence than previous methods.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121756671","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We construct a Bayesian model that integrates topdown with bottom-up criteria, capitalizing on their relative merits to obtain figure-ground segmentation that is shape-specific and texture invariant. A hierarchy of bottom-up segments in multiple scales is used to construct a prior on all possible figure-ground segmentations of the image. This prior is used by our top-down part to query and detect object parts in the image using stored shape templates. The detected parts are integrated to produce a global approximation for the object’s shape, which is then used by an inference algorithm to produce the final segmentation. Experiments with a large sample of horse and runner images demonstrate strong figure-ground segmentation despite high object and background variability. The segmentations are robust to changes in appearance since the matching component depends on shape criteria alone. The model may be useful for additional visual tasks requiring labeling, such as the segmentation of multiple scene objects.
{"title":"Shape Guided Object Segmentation","authors":"Eran Borenstein, Jitendra Malik","doi":"10.1109/CVPR.2006.276","DOIUrl":"https://doi.org/10.1109/CVPR.2006.276","url":null,"abstract":"We construct a Bayesian model that integrates topdown with bottom-up criteria, capitalizing on their relative merits to obtain figure-ground segmentation that is shape-specific and texture invariant. A hierarchy of bottom-up segments in multiple scales is used to construct a prior on all possible figure-ground segmentations of the image. This prior is used by our top-down part to query and detect object parts in the image using stored shape templates. The detected parts are integrated to produce a global approximation for the object’s shape, which is then used by an inference algorithm to produce the final segmentation. Experiments with a large sample of horse and runner images demonstrate strong figure-ground segmentation despite high object and background variability. The segmentations are robust to changes in appearance since the matching component depends on shape criteria alone. The model may be useful for additional visual tasks requiring labeling, such as the segmentation of multiple scene objects.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121876858","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents a 3D approach for recognizing faces based on Principal Component Analysis (PCA). The approach addresses the issue of proper 3D face alignment required by PCA for maximum data compression and good generalization performance for new untrained faces. This issue has traditionally been addressed by 2D data normalization, a step that eliminates 3D object size information important for the recognition process. We achieve correspondence of facial points by registering a 3D face to a scaled generic 3D reference face and subsequently perform a surface normal search algorithm. 3D scaling of the generic reference face is performed to enable better alignment of facial points while preserving important 3D size information in the input face. The benefits of this approach for 3D face recognition and dimensionality reduction have been demonstrated on components of the Face Recognition Grand Challenge (FRGC) database versions 1 and 2.
{"title":"3D Face Recognition Using 3D Alignment for PCA","authors":"T. Russ, Chris Boehnen, Tanya Peters","doi":"10.1109/CVPR.2006.13","DOIUrl":"https://doi.org/10.1109/CVPR.2006.13","url":null,"abstract":"This paper presents a 3D approach for recognizing faces based on Principal Component Analysis (PCA). The approach addresses the issue of proper 3D face alignment required by PCA for maximum data compression and good generalization performance for new untrained faces. This issue has traditionally been addressed by 2D data normalization, a step that eliminates 3D object size information important for the recognition process. We achieve correspondence of facial points by registering a 3D face to a scaled generic 3D reference face and subsequently perform a surface normal search algorithm. 3D scaling of the generic reference face is performed to enable better alignment of facial points while preserving important 3D size information in the input face. The benefits of this approach for 3D face recognition and dimensionality reduction have been demonstrated on components of the Face Recognition Grand Challenge (FRGC) database versions 1 and 2.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"247 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121877759","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the growing interest in object categorization various methods have emerged that perform well in this challenging task, yet are inherently limited to only a moderate number of object classes. In pursuit of a more general categorization system this paper proposes a way to overcome the computational complexity encompassing the enormous number of different object categories by exploiting the statistical properties of the highly structured visual world. Our approach proposes a hierarchical acquisition of generic parts of object structure, varying from simple to more complex ones, which stem from the favorable statistics of natural images. The parts recovered in the individual layers of the hierarchy can be used in a top-down manner resulting in a robust statistical engine that could be efficiently used within many of the current categorization systems. The proposed approach has been applied to large image datasets yielding important statistical insights into the generic parts of object structure.
{"title":"Hierarchical Statistical Learning of Generic Parts of Object Structure","authors":"S. Fidler, Gregor Berginc, A. Leonardis","doi":"10.1109/CVPR.2006.134","DOIUrl":"https://doi.org/10.1109/CVPR.2006.134","url":null,"abstract":"With the growing interest in object categorization various methods have emerged that perform well in this challenging task, yet are inherently limited to only a moderate number of object classes. In pursuit of a more general categorization system this paper proposes a way to overcome the computational complexity encompassing the enormous number of different object categories by exploiting the statistical properties of the highly structured visual world. Our approach proposes a hierarchical acquisition of generic parts of object structure, varying from simple to more complex ones, which stem from the favorable statistics of natural images. The parts recovered in the individual layers of the hierarchy can be used in a top-down manner resulting in a robust statistical engine that could be efficiently used within many of the current categorization systems. The proposed approach has been applied to large image datasets yielding important statistical insights into the generic parts of object structure.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"185 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132587234","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
V. Kovalev, N. Harder, B. Neumann, Michael Held, U. Liebel, H. Erfle, J. Ellenberg, R. Eils, K. Rohr
We investigate different approaches for efficient feature space reduction and compare different methods for cell classification. The application context is the development of automatic methods for analysing fluorescence microscopy images with the goal to identify those genes that are involved in the mitosis of human cells (cell division). We distinguish four cell classes comprising interphase cells, mitotic cells, apoptotic cells, and cells with clustered nuclei. Feature space reduction was performed using the Principal Component Analysis and Independent Component Analysis methods. Six classification methods were examined including unsupervised clustering algorithms such as K-means, Hard Competitive Learning, and Neural Gas as well as Hierarchical Clustering, Support Vector Machines, and Random Forests classifiers. Detailed results on the cell image classification accuracy and computational efficiency achieved using different feature sets and different classification methods are reported.
{"title":"Feature Selection for Evaluating Fluorescence Microscopy Images in Genome-Wide Cell Screens","authors":"V. Kovalev, N. Harder, B. Neumann, Michael Held, U. Liebel, H. Erfle, J. Ellenberg, R. Eils, K. Rohr","doi":"10.1109/CVPR.2006.121","DOIUrl":"https://doi.org/10.1109/CVPR.2006.121","url":null,"abstract":"We investigate different approaches for efficient feature space reduction and compare different methods for cell classification. The application context is the development of automatic methods for analysing fluorescence microscopy images with the goal to identify those genes that are involved in the mitosis of human cells (cell division). We distinguish four cell classes comprising interphase cells, mitotic cells, apoptotic cells, and cells with clustered nuclei. Feature space reduction was performed using the Principal Component Analysis and Independent Component Analysis methods. Six classification methods were examined including unsupervised clustering algorithms such as K-means, Hard Competitive Learning, and Neural Gas as well as Hierarchical Clustering, Support Vector Machines, and Random Forests classifiers. Detailed results on the cell image classification accuracy and computational efficiency achieved using different feature sets and different classification methods are reported.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132738351","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We consider visual category recognition in the framework of measuring similarities, or equivalently perceptual distances, to prototype examples of categories. This approach is quite flexible, and permits recognition based on color, texture, and particularly shape, in a homogeneous framework. While nearest neighbor classifiers are natural in this setting, they suffer from the problem of high variance (in bias-variance decomposition) in the case of limited sampling. Alternatively, one could use support vector machines but they involve time-consuming optimization and computation of pairwise distances. We propose a hybrid of these two methods which deals naturally with the multiclass setting, has reasonable computational complexity both in training and at run time, and yields excellent results in practice. The basic idea is to find close neighbors to a query sample and train a local support vector machine that preserves the distance function on the collection of neighbors. Our method can be applied to large, multiclass data sets for which it outperforms nearest neighbor and support vector machines, and remains efficient when the problem becomes intractable for support vector machines. A wide variety of distance functions can be used and our experiments show state-of-the-art performance on a number of benchmark data sets for shape and texture classification (MNIST, USPS, CUReT) and object recognition (Caltech- 101). On Caltech-101 we achieved a correct classification rate of 59.05%(±0.56%) at 15 training images per class, and 66.23%(±0.48%) at 30 training images.
{"title":"SVM-KNN: Discriminative Nearest Neighbor Classification for Visual Category Recognition","authors":"Haotong Zhang, A. Berg, M. Maire, Jitendra Malik","doi":"10.1109/CVPR.2006.301","DOIUrl":"https://doi.org/10.1109/CVPR.2006.301","url":null,"abstract":"We consider visual category recognition in the framework of measuring similarities, or equivalently perceptual distances, to prototype examples of categories. This approach is quite flexible, and permits recognition based on color, texture, and particularly shape, in a homogeneous framework. While nearest neighbor classifiers are natural in this setting, they suffer from the problem of high variance (in bias-variance decomposition) in the case of limited sampling. Alternatively, one could use support vector machines but they involve time-consuming optimization and computation of pairwise distances. We propose a hybrid of these two methods which deals naturally with the multiclass setting, has reasonable computational complexity both in training and at run time, and yields excellent results in practice. The basic idea is to find close neighbors to a query sample and train a local support vector machine that preserves the distance function on the collection of neighbors. Our method can be applied to large, multiclass data sets for which it outperforms nearest neighbor and support vector machines, and remains efficient when the problem becomes intractable for support vector machines. A wide variety of distance functions can be used and our experiments show state-of-the-art performance on a number of benchmark data sets for shape and texture classification (MNIST, USPS, CUReT) and object recognition (Caltech- 101). On Caltech-101 we achieved a correct classification rate of 59.05%(±0.56%) at 15 training images per class, and 66.23%(±0.48%) at 30 training images.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132812344","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Multi-camera tracking systems often must maintain consistent identity labels of the targets across views to recover 3D trajectories and fully take advantage of the additional information available from the multiple sensors. Previous approaches to the "correspondence across views" problem include matching features, using camera calibration information, and computing homographies between views under the assumption that the world is planar. However, it can be difficult to match features across significantly different views. Furthermore, calibration information is not always available and planar world hypothesis can be too restrictive. In this paper, a new approach is presented for matching correspondences based on the use of nonlinear manifold learning and system dynamics identification. The proposed approach does not require similar views, calibration nor geometric assumptions of the 3D environment, and is robust to noise and occlusion. Experimental results demonstrate the use of this approach to generate and predict views in cases where identity labels become ambiguous.
{"title":"Modeling Correspondences for Multi-Camera Tracking Using Nonlinear Manifold Learning and Target Dynamics","authors":"Vlad I. Morariu, O. Camps","doi":"10.1109/CVPR.2006.189","DOIUrl":"https://doi.org/10.1109/CVPR.2006.189","url":null,"abstract":"Multi-camera tracking systems often must maintain consistent identity labels of the targets across views to recover 3D trajectories and fully take advantage of the additional information available from the multiple sensors. Previous approaches to the \"correspondence across views\" problem include matching features, using camera calibration information, and computing homographies between views under the assumption that the world is planar. However, it can be difficult to match features across significantly different views. Furthermore, calibration information is not always available and planar world hypothesis can be too restrictive. In this paper, a new approach is presented for matching correspondences based on the use of nonlinear manifold learning and system dynamics identification. The proposed approach does not require similar views, calibration nor geometric assumptions of the 3D environment, and is robust to noise and occlusion. Experimental results demonstrate the use of this approach to generate and predict views in cases where identity labels become ambiguous.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132936233","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Many computer vision applications, such as scene analysis and medical image interpretation, are ill-suited for traditional classification where each image can only be associated with a single class. This has stimulated recent work in multi-label learning where a given image can be tagged with multiple class labels. A serious problem with existing approaches is that they are unable to exploit correlations between class labels. This paper presents a novel framework for multi-label learning termed Correlated Label Propagation (CLP) that explicitly models interactions between labels in an efficient manner. As in standard label propagation, labels attached to training data points are propagated to test data points; however, unlike standard algorithms that treat each label independently, CLP simultaneously co-propagates multiple labels. Existing work eschews such an approach since naive algorithms for label co-propagation are intractable. We present an algorithm based on properties of submodular functions that efficiently finds an optimal solution. Our experiments demonstrate that CLP leads to significant gains in precision/recall against standard techniques on two real-world computer vision tasks involving several hundred labels.
{"title":"Correlated Label Propagation with Application to Multi-label Learning","authors":"Feng Kang, Rong Jin, R. Sukthankar","doi":"10.1109/CVPR.2006.90","DOIUrl":"https://doi.org/10.1109/CVPR.2006.90","url":null,"abstract":"Many computer vision applications, such as scene analysis and medical image interpretation, are ill-suited for traditional classification where each image can only be associated with a single class. This has stimulated recent work in multi-label learning where a given image can be tagged with multiple class labels. A serious problem with existing approaches is that they are unable to exploit correlations between class labels. This paper presents a novel framework for multi-label learning termed Correlated Label Propagation (CLP) that explicitly models interactions between labels in an efficient manner. As in standard label propagation, labels attached to training data points are propagated to test data points; however, unlike standard algorithms that treat each label independently, CLP simultaneously co-propagates multiple labels. Existing work eschews such an approach since naive algorithms for label co-propagation are intractable. We present an algorithm based on properties of submodular functions that efficiently finds an optimal solution. Our experiments demonstrate that CLP leads to significant gains in precision/recall against standard techniques on two real-world computer vision tasks involving several hundred labels.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133914854","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Erik B. Sudderth, A. Torralba, W. Freeman, A. Willsky
We develop an integrated, probabilistic model for the appearance and three-dimensional geometry of cluttered scenes. Object categories are modeled via distributions over the 3D location and appearance of visual features. Uncertainty in the number of object instances depicted in a particular image is then achieved via a transformed Dirichlet process. In contrast with image-based approaches to object recognition, we model scale variations as the perspective projection of objects in different 3D poses. To calibrate the underlying geometry, we incorporate binocular stereo images into the training process. A robust likelihood model accounts for outliers in matched stereo features, allowing effective learning of 3D object structure from partial 2D segmentations. Applied to a dataset of office scenes, our model detects objects at multiple scales via a coarse reconstruction of the corresponding 3D geometry.
{"title":"Depth from Familiar Objects: A Hierarchical Model for 3D Scenes","authors":"Erik B. Sudderth, A. Torralba, W. Freeman, A. Willsky","doi":"10.1109/CVPR.2006.97","DOIUrl":"https://doi.org/10.1109/CVPR.2006.97","url":null,"abstract":"We develop an integrated, probabilistic model for the appearance and three-dimensional geometry of cluttered scenes. Object categories are modeled via distributions over the 3D location and appearance of visual features. Uncertainty in the number of object instances depicted in a particular image is then achieved via a transformed Dirichlet process. In contrast with image-based approaches to object recognition, we model scale variations as the perspective projection of objects in different 3D poses. To calibrate the underlying geometry, we incorporate binocular stereo images into the training process. A robust likelihood model accounts for outliers in matched stereo features, allowing effective learning of 3D object structure from partial 2D segmentations. Applied to a dataset of office scenes, our model detects objects at multiple scales via a coarse reconstruction of the corresponding 3D geometry.","PeriodicalId":421737,"journal":{"name":"2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2006-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115047964","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：