3D Object Detection and Viewpoint Estimation with a Deformable 3D Cuboid Model

By Sanja Fidler, Sven J. Dickinson, Raquel Urtasun. 2012

199 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, 3D Vision & Pose

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In 3D Vision & Pose

Citations in the corpus

On the Transfer of Disentangled Representations in Realistic Settings (2020, background): “Cars3D (Reed et al., 2015) exhibits diﬀerent car models from Fidler et al. (2012) under diﬀerent camera viewpoints.”
On the Transfer of Inductive Bias from Simulation to the Real World: a New Disentanglement Dataset (2019, methodology): “For Cars3D,199 CAD models from [11] were used to generate 64x64 color renderings from 24 rotation angles each offset by 15 degrees [41] and available at http://www.”
Cross-View Action Modeling, Learning, and Recognition (2014, methodology): “In addition, the 2D DPM model can be extended to 3D [5, 12] to facilitate multiActions”
Voting for Voting in Online Point Cloud Object Detection (2015, background): “A body of work in monocular object detection with 3D pose estimation [9, 7] also characterises objects by 3D bounding boxes identified by sliding a window in 3D.”
Pix3D: Dataset and Methods for Single-Image 3D Shape Modeling (2018, methodology): “Researchers have also proposed numerous methods on 3D pose estimation [13, 42, 55, 58].”
Single Image 3D Interpreter Network (2016, methodology): “Some previous methods formulated it as a classiﬁcation or regression problem, and aimed to directly estimate the viewpoint from an image [35,11].”
Monocular Object Instance Segmentation and Depth Ordering with CNNs (2015, methodology): “It solves a submodular energy that makes use of ground-truth 3D bounding box annotations, point cloud, stereo imagery and shape priors from rendered CAD models [5] in order to generate a pixel-wise labeling of the object.”
ALICE: Towards Understanding Adversarial Learning for Joint Distribution Matching (2017, background): “(i) Car-to-Car [24]: each domain (x and z) includes car images in 11 different angles, on which we seek to demonstrate the power of adversarially learned reconstruction and weak supervision.”
Instance-Level Segmentation for Autonomous Driving with Deep Densely Connected MRFs (2015, methodology): “Since KITTI does not have instance-level segmentations, we use [5, 7] to obtain the segmentations for our training set.”
Unsupervised Visual Attribute Transfer with Reconfigurable Generative Adversarial Networks (2017, methodology): “To investigate whether our model can cope with this, we combined two rendered 3D image datasets: 3D Car [20] and 3D Face [21].”
Joint SFM and detection cues for monocular 3D localization in road scenes (2015, background): “Also related are works on 3D object detection in a single image [6, 14].”
Beat the MTurkers: Automatic Image Labeling from Weak 3D Supervision (2014, methodology): “Instead, we obtain this information from the car manufacturers and other online resources, as each CAD model’s meta-data contains information about the car brand and model [8].”
ObjectNet3D: A Large Scale Database for 3D Object Recognition (2016, background): “[9,24,42,33,15,39]), as the field has benefited from the introduction of several important databases that provide 3D annotations to 2D objects.”
One-Sided Unsupervised Domain Mapping (2017, background): “The 3D car dataset Fidler et al. (2012) consists of rendered images of 3D cars whose degree varies at 15◦ intervals.”
Mesh R-CNN (2019, background): “Other approaches predict novel 3D shapes as sets of 3D points [8, 34], patches [15, 70], or geometric primitives [9, 64, 67]; others use deep networks to model signed distance functions [42].”
From Points to Parts: 3D Object Detection From Point Cloud With Part-Aware and Part-Aggregation Network (2019, background): “[54], [55], [56] extended the DPM to 3D world to reason the parts in 3D and estimate the object poses, where [54] modeled the object as a 3D cuboid with both deformable faces and deformable parts, [55] proposed a 3D DPM that generates a full 3D object model with continuous appearance…”