Multiview feature distributions for object detection and continuous pose estimation

Abstract

This paper presents a multiview model of object categories, generally applicable to virtually any type of image features, and methods to efficiently perform, in a unified manner, detection, localization and continuous pose estimation in novel scenes. We represent appearance as distributions of low-level, fine-grained image features. Multiview models encode the appearance of objects at discrete viewpoints, and, in addition, how these viewpoints deform into one another as the viewpoint continuously varies (as detected from optical flow between training examples). Using a measure of similarity between an arbitrary test image and such a model at chosen viewpoints, we perform all tasks mentioned above with a common method. We leverage the simplicity of low-level image features, such as points extracted along edges, or coarse-scale gradients extracted densely over the images, by building probabilistic templates, i.e. distributions of features, learned from one or several training examples. We efficiently handle these distributions with probabilistic techniques such as kernel density estimation, Monte Carlo integration and importance sampling. We provide an extensive evaluation on a wide variety of benchmark datasets. We demonstrate performance on the "ETHZ Shape" dataset, with single (hand-drawn) and multiple training examples, well above baseline methods, on par with a number of more task-specific methods. We obtain remarkable performance on the recognition of more complex objects, notably the cars of the "3D Object" dataset of Savarese et al. with detection rates of 92.5% and an accuracy in pose estimation of 91%. We perform better than the state-of-the-art on continuous pose estimation with the "rotating cars" dataset of Ozuysal et al. . We also demonstrate particular capabilities with a novel dataset featuring non-textured objects of undistinctive shapes, the pose of which can only be determined from shading, captured here by coarse scale intensity gradients.

Learning Semantics of Gestural Instructions for Human-Robot Collaboration

“…We use our probabilistic appearance-based pose estimation (PAPE) method (Erkent et al, 2016 ; Shukla et al, 2016 ) to detect hand gestures. It is based on probabilistic representations of objects and scenes by Teney and Piater ( 2014 ).…”

Continuous Pose Estimation with a Spatial Ensemble of Fisher Regressors

“…In [1], a complex object-viewpoint manifold is built and then untangled by factorizing the manifold in a view-invariant category representation and a category-invariant viewpoint representation, where the latter is used for pose estimation. In [27], the class representation is a probability distribution depending on the image and pose coordinates of extracted edge features. The object pose in the query image is estimated by marginalizing out the product of the query distribution and the class distribution with respect to the spatial coordinates.…”

“…Glasner et al [11] // 24.8 Pepik et al [21] // 4.7 Ozuysal et al [20] 46.48 // Redondo et al [24] 39.8 7 Teney et al [27] 34.7 5.2 Torki et al [28] tional experiment on the EPFL dataset. Instead of the term described in Equation ( 12), we replace it with a uniform distribution over [0 • , 360 • ).…”

Theoretical accuracy assessment of model-based photogrammetric approach for pose estimation of cylindrical elements

“…Both for single view (Shiu and Huang 1991b, Ferri et al 1993, Puech et al 1997, Penman and Alwesh 2006, Doignon and De Mathelin 2007, Liu and Hu 2014 approaches or multi-view ones (Houqin and Jianbo 2008, Becke 2015, Teney and Piater 2014, Becke and Schlegl 2015, Zhu et al 2015, Zhang et al 2017, cylinder pose estimation can be estimated based on several approaches or on a combination of them. Depending on the image data taken into consideration, only cylinder orientation in three degrees of freedom (dof) or five dof pose can be established employing the contour data corresponding to cylinder's circular borders.…”