Efficient 3D Scene Labeling Using Fields of Trees

Kähler, Olaf; Reid, Ian

doi:10.1109/iccv.2013.380

Cited by 30 publications

(46 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the vehicle part domain most of the parts are either clearly vertical or horizontal. Finally we include the mean, median, and max of the plane fit errors of the points in each part, the three eigenvalues from the plane estimation (λ 1 , λ 2 , λ 3 in descending order), and the differences between adjacent eigenvalues that have been referred to as linearity (λ 1 − λ 2 ) and planarity (λ 2 − λ 3 ) in previous work [30], [31].…”

Section: Part-level Feature Extractionmentioning

confidence: 99%

Classification of Vehicle Parts in Unstructured 3D Point Clouds

Zelener¹,

Mordohai

Stamos³

2014

2014 2nd International Conference on 3D Vision

View full text Add to dashboard Cite

Abstract-Unprecedented amounts of 3D data can be acquired in urban environments, but their use for scene understanding is challenging due to varying data resolution and variability of objects in the same class. An additional challenge is due to the nature of the point clouds themselves, since they lack detailed geometric or semantic information that would aid scene understanding. In this paper we present a general algorithm for segmenting and jointly classifying object parts and the object itself. Our pipeline consists of local feature extraction, robust RANSAC part segmentation, partlevel feature extraction, a structured model for parts in objects, and classification using state-of-the-art classifiers. We have tested this pipeline in a very challenging dataset that consists of real world scans of vehicles. Our contributions include the development of a segmentation and classification pipeline for objects and their parts; and a method for segmentation that is robust to the complexity of unstructured 3D points clouds, as well as a part ordering strategy for the sequential structured model and a joint feature representation between object parts.

show abstract

Section: Part-level Feature Extractionmentioning

confidence: 99%

Classification of Vehicle Parts in Unstructured 3D Point Clouds

Zelener¹,

Mordohai

Stamos³

2014

2014 2nd International Conference on 3D Vision

View full text Add to dashboard Cite

show abstract

“…Although there has been remarkable progress in 2D image semantic segmentation, 3D scene semantic labelling has not received as much attention or achievement (though some notable exceptions exist, e.g., [12,8]). Unlike 2D images which capture specific views, 3D reconstructed point clouds cover the whole scene with a large number of things and stuff, making label prediction more challenging.…”

Section: Introductionmentioning

confidence: 99%

“…The model parameters are learned from training data using a Structured Support Vector Machine (SSVM) approach. Kahler and Reid [8] learn similar potential functions using Decision Tree Fields (DTF) [19] and Regression Tree Fields (RTF) [7]. The tree based method [8] is more efficient and yields similar segmentation accuracies.…”

Section: Introductionmentioning

confidence: 99%

“…Kahler and Reid [8] learn similar potential functions using Decision Tree Fields (DTF) [19] and Regression Tree Fields (RTF) [7]. The tree based method [8] is more efficient and yields similar segmentation accuracies. Although initial promising results have been reported, both of these models restrict themselves to pairwise potential functions limiting their ability to encode complex geometrical and topological characteristics of 3D scenes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Hierarchical Higher-Order Regression Forest Fields: An Application to 3D Indoor Scene Labelling

Pham

Reid

Latif

et al. 2015

2015 IEEE International Conference on Computer Vision (ICCV)

Self Cite

View full text Add to dashboard Cite

This paper addresses the problem of semantic segmentation of 3D indoor scenes reconstructed from RGB-D images. Traditionally label prediction for 3D points is tackled by employing graphical models that capture scene features and complex relations between different class labels. However, the existing work is restricted to pairwise conditional random fields, which are insufficient when encoding rich scene context. In this work we propose models with higher-order potentials to describe complex relational information from the 3D scenes. Specifically, we relax the labelling problem to a regression, and generalize the higher-order associative P n Potts model to a new family of arbitrary higherorder models based on regression forests. We show that these models, like the robust P n models, can still be decomposed into the sum of pairwise terms by introducing auxiliary variables. Moreover, our proposed higher-order models also permit extension to hierarchical random fields, which allows for the integration of scene context and features computed at different scales. Our potential functions are constructed based on regression forests encoding Gaussian densities that admit efficient inference. The parameters of our model are learned from training data using a structured learning approach. Results on two datasets show clear improvements over current state-of-the-art methods.

show abstract

“…Zhu et al [32] provide a good survey of semantic segmentation methods using RGB images. The use of RGBD images has a shorter history but a significant amount of works have already been 1 presented [13,10,26,3]. RGBD images carry more information but depth maps can be noisy and may contain large areas with missing measurements.…”

Section: Object and Materials Attribute Detectionmentioning

confidence: 99%

Room Layout Estimation with Object and Material Attributes Information Using a Spherical Camera

Kim

Campos

Hilton

2016

2016 Fourth International Conference on 3D Vision (3DV)

View full text Add to dashboard Cite

In this paper we propose a pipeline for estimating 3D room layout with object and material attribute prediction using a spherical stereo image pair. We assume that the room and objects can be represented as cuboids aligned to the main axes of the room coordinate (Manhattan world). A spherical stereo alignment algorithm is proposed to align two spherical images to the global world coordinate system. Depth information of the scene is estimated by stereo matching between images. Cubic projection images of the spherical RGB and estimated depth are used for object and material attribute detection. A single Convolutional Neural Network is designed to assign object and attribute labels to geometrical elements built from the spherical image. Finally simplified room layout is reconstructed by cuboid fitting. The reconstructed cuboid-based model shows the structure of the scene with object information and material attributes.

show abstract

Efficient 3D Scene Labeling Using Fields of Trees

Cited by 30 publications

References 13 publications

Classification of Vehicle Parts in Unstructured 3D Point Clouds

Classification of Vehicle Parts in Unstructured 3D Point Clouds

Hierarchical Higher-Order Regression Forest Fields: An Application to 3D Indoor Scene Labelling

Room Layout Estimation with Object and Material Attributes Information Using a Spherical Camera

Contact Info

Product

Resources

About