Teaching 3D geometry to deformable part models

By Bojan Pepik, Michael Stark, Peter Gehler, Bernt Schiele. 2012

243 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

End-to-End Training of Deep Visuomotor Policies (2015, methodology): “…window to localize the object, reducing the task to classiﬁcation [8], perform regression to a heatmap of manually labeled keypoints [40], requiring precise knowledge of the object position in the image and camera calibration, or use 3D models to localize previously scanned objects [30, 36].”
Single Image 3D Interpreter Network (2016, methodology): “VDPM [17] and DPM-VOC+VP [50] are trained on PASCAL VOC 2012, V&K [18] is trained on PASCAL 3D+, Su et al .”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, methodology): “mount of annotated data. Synthetic data based adaptation. Early methods use 3D CAD models to improve solutions for pose and viewpoint estimation [161, 162, 163, 164], object and object part detection [165, 166, 167, 168, 169, 170, 171, 172], segmentation and scene understanding [173, 174, 175]. The ”
VirtualWorlds as Proxy for Multi-object Tracking Analysis (2016, background): “Several works investigate the use of 3D synthetic data to tackle standard 2D computer vision problems such as object detection [5], face recognition, scene understanding [6], and optical flow estimation [7].”
Learning 3D Object Categories by Looking Around Them (2017, methodology): “Alternatively, CAD models can be used to train a network to directly predict the 3D shape of an object [10, 41, 38, 7].”
Beyond PASCAL: A benchmark for 3D object detection in the wild (2014, methodology): “DPM [6] -16V 38.0 / 15.9 45.6 / 22.9 0.7 / 0.3 – / – 55.3 / 49.0 46.0 / 29.6 10.2 / 6.1 6.2 / 2.3 38.1 / 16.7 11.8 / 7.1 28.5 / 20.2 30.7 / 19.9 28.3 / 17.3 DPM-VOC+VP [21] The results of DPM, VDPM and DPM-VOC+VP are shown.”
Modeling Occlusion by Discriminative AND-OR Structures (2013, methodology): “While several recent methods use the synthesized images to learn part geometry [24] or detailed part appearance [20], our method only uses them to learn the configuration structure of occlusion, where the appearance terms are still trained using real images.”
Holistic Scene Understanding for 3D Object Detection with RGBD Cameras (2013, background): “Over the last decade a variety of approaches have been developed in order to infer 3D objects from monocular imagery [12] [6] [19].”
Sim4CV: A Photo-Realistic Simulator for Computer Vision Applications (2017, background): “A broad range of work has recently exploited physics based simulators for learning purposes, namely in animation and motion planning ([18; 15; 27; 24; 51; 16; 14]), scene understanding ([3; 38]), pedestrian detection ([29]), and identiﬁcation of 2D/3D objects ([17; 31; 39]).”
6-DoF object pose from semantic keypoints (2017, methodology): “To address class variability and viewpoint, various approaches have used a collection of 2D appearance-based part templates trained separately on discretized views [1], [13], [14], [15], [16].”
A General Pipeline for 3D Detection of Vehicles (2018, background): “Deformable part-based models (DPM) can be found in [25], [26], [27] and [28].”
Fast Single Shot Detection and Pose Estimation (2016, methodology): “[11]; they formulate the joint object detection and pose estimation problem as a structured prediction problem.”
Deep MANTA: A Coarse-to-Fine Many-Task Network for Joint 2D and 3D Vehicle Analysis from Monocular Image (2017, background): “They generally divide the viewing sphere in several bins to learn multi-class models where each bin corresponds to a class [27, 41, 47, 22, 29].”
Detailed 3D Representations for Object Recognition and Modeling (2013, methodology): “In particular, we use the recently proposed multiview extension of the deformable part model by Pepik et al. [45] , which has been shown to yield excellent performance w.r.t. both 2D bounding box localization and coarse viewpoint classification.”
Articulation-Aware Canonical Surface Mapping (2020, background): “The task of estimating pose for rigid objects has been central to understanding objects in 3D scenes, and addressed by several works over the decades, from matching based methods [12, 27, 37], to recent CNN based predictors [29, 32].”
Learning descriptors for object recognition and 3D pose estimation (2015, background): “[26] extends the Deformable Part Model to 3D object detection and pose estimation.”