Completing 3D object shape from one depth image

By Jason Rock, Tanmay Gupta, Justin Thorsen, JunYoung Gwak, Daeyun Shin, Derek Hoiem. 2015

176 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computational Mechanics, 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 Aerospace Engineering

In Computational Mechanics

In Computer Vision and Pattern Recognition

In 3D Vision & Pose

Citations in the corpus

DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation (2019, methodology): “Most of these methods adopt encoder-decoder architectures to reduce partial inputs of occupancy voxels [56], discrete SDF voxels [16], depth maps [44], RGB images [14, 55] or point clouds [49] into a latent vector and subsequently generate a prediction of full volumetric shape based on learned prior”
Learning to Reconstruct Shapes from Unseen Classes (2018, background): “Very recently, researchers have begun to look at the generalization power of 3D reconstruction algorithms [Shin et al., 2018, Jayaraman et al., 2018, Rock et al., 2015, Funk and Liu, 2017].”
MarrNet: 3D Shape Reconstruction via 2.5D Sketches (2017, methodology): “This idea is also widely used in 3D shape completion from depths or silhouettes [Firman et al., 2016, Rock et al., 2015, Dai et al., 2017].”
Shape Completion Using 3D-Encoder-Predictor CNNs and Shape Synthesis (2016, methodology): “ShapeRe-con [32] performs slightly better, but overall, it is heavily dependent on ﬁnding good nearest neighbors; the available implementation was also trained only on a subset of classes.”
3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction (2016, background): “The shape priors are typically encoded in the form of simple 3D primitives as demonstrated by early pioneering works [19,20] or learned from rich repositories of 3D CAD models [21,22,23], whereby the concept of ﬁtting 3D models to images of faces was explored to a much larger extent [24,25,26].”
Semantic Scene Completion from a Single Depth Image (2016, background): “Other prior works focus on single object shape completion [33, 28, 34, 32].”
Deformable Shape Completion with Graph Convolutional Autoencoders (2017, background): “The application addressed in this paper is a very active research area in computer vision and graphics, ranging from completion of small holes [48] and larger missing regions in individual objects [1, 45, 55, 61, 49], to entire scenes [51].”
Learning 3D Object Categories by Looking Around Them (2017, methodology): “We evaluate the point cloud completion module of VpDR-Net by comparing ground truth point clouds C to the point clouds Ĉ predicted by Φpcl using: (1) the voxel intersection-over-union (VIoU) measure that computes the Jaccard similarity between the volumetric representations of Ĉ and C, and (2) the n”
Pixels, Voxels, and Views: A Study of Shape Representations for Single View 3D Object Shape Prediction (2018, methodology): “Volumetric shape representations: Several recent studies offer methods to generate volumetric object shapes from one or a few images [20, 7, 13, 3, 22, 19].”
SemanticKITTI: A Dataset for Semantic Scene Understanding of LiDAR Sequences (2019, methodology): “In case of NYUv2 [48], CAD models were fit into the scene [45] using an RGB-D image captured by a Kinect sensor.”
Shape Completion from a Single RGBD Image (2017, methodology): “The method of [23] is directly adopted to incorporate the initialization, visibility and symmetry constraints.”
High-Resolution Shape Completion Using Deep Neural Networks for Global Structure and Local Geometry Inference (2017, background): “One simple approach is to retrieve the most similar shape from a database and use it as a template that can be deformed to ﬁt the input raw data ( [19] [21]).”
SceneNN: A Scene Meshes Dataset with aNNotations (2016, methodology): “[22] proposed to complete an object by retrieving a similar depth map from a dataset, and then find its corresponding 3D model and deforming the model to fit the input scan.”
Anisotropic Convolutional Networks for 3D Semantic Scene Completion (2020, methodology): “Following SSCNet [16], we use the 3D annotations provided by [14] for semantic scene completion task.”
3D Object Dense Reconstruction from a Single Depth View (2018, unknown): “Given a partial shape input, [31] [32] [33] [34] [35] [36] try to retrieve an identical or most likely 3D model and align it with the partial scan.”
Shape completion enabled robotic grasping (2016, methodology): “In addition, work by [7] uses an exemplar based approach for the same task.”
Structured Prediction of Unobserved Voxels from a Single Depth Image (2016, methodology): “Deformation based methods such as [4, 44] directly deform a target mesh to the observed data but can fail when an incorrect model is retrieved from the database.”
Weakly Supervised 3D Reconstruction with Adversarial Constraint (2017, methodology): “Among many lines of work within the 3D reconstruction [18, 25, 13, 14, 3, 8, 23, 38, 49, 35], ours is related to recent works that use neural networks for 3D voxel reconstruction.”