Recovering the spatial layout of cluttered rooms

By Varsha Hedau, Derek Hoiem, D.A. Forsyth. IEEE International Conference on Computer Vision, 2009

557 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/451a06626afe8dd70099c7dfec86de7af909a062

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

RGB-(D) scene labeling: Features and algorithms (2012, background): “Related but different are the works on indoor scene recognition [25] and Manhattan world box labeling [20] [11] have been popular in scene labeling [8, 18, 16], which use boosting to transform the outputs from a multi-scale texton filterbank.”
Support Surface Prediction in Indoor Scenes (2013, background): “One line of work is to predict boxy layouts from RGB images [6, 7, 10, 8, 5, 12, 11], in which occlusion and lack of 3D sensors is combatted with simple scene models and strong priors.”
Marr Revisited: 2D-3D Alignment via Surface Normal Prediction (2016, background): “Prior work has sought to recover the overall global shape of the room by fitting a global parametric 3D box [17, 39] or recovering informative edge maps [29] that align to the shape of the room, typically based on Manhattan world constraints [8, 21].”
Patch to the Future: Unsupervised Visual Prediction (2014, background): “While prediction has been extensively studied in the realm of biological vision [1], in the field of computational vision, most of the research focus has been on understanding and inferring semantic [7, 27, 32] and geometric [10, 14] knowledge of what can be seen in an image or a video.”
Estimating the 3D Layout of Indoor Scenes and Its Clutter from Depth Sensors (2013, methodology): “Indoor scene understanding approaches have taken advantage of the Manhattan world properties of rooms and frame the layout estimation task as the prediction of a 3D cuboid aligned with the three main dominant orientations [10, 11, 18, 22, 23, 16, 29].”
Cross-View Semantic Segmentation for Sensing Surroundings (2019, methodology): “ and explore the corresponding cross-view segmentation network architectures. Layout estimation and view synthesis. Estimating layout has been an active topic of research (i.e. room layout estimation [50, 22, 43, 19, 16], free space estimation [17]). Most of the previous methods use annotations of t”
IM2CAD (2016, methodology): “Computer vision algorithms have also become increasingly good at this task in the last few years by following a paradigm introduced by Hedau et al. [15] and Lee et al. [22] in which a set of room shapes are hypothesized (typically 3D boxes), and evaluated using features in the image.”
Three-Dimensional Object Detection and Layout Prediction Using Clouds of Oriented Gradients (2016, methodology): “[12] use a structured model to rerank layout hypotheses, Schwing et al.”
3DNN: 3D Nearest Neighbor (2014, methodology): “We show this by building upon the work of Hedau et al. (2009) and by demonstrating how prior 3D models can be integrated with this existing approach for room layout estimation to help discover the identity, locations and orientations of objects from a single image.”
Single Image 3D without a Single 3D Image (2015, methodology): “We use the Manhattan-world assumption [8] to reduce our label space to three orthogonal directions; we find these directions and rectification homographies for them using vanishing points estimated by [20].”
Estimating Spatial Layout of Rooms using Volumetric Reasoning about Objects and Surfaces (2010, methodology): “We evaluated our 3D geometric reasoning approach on an indoor image dataset introduced in [12].”
Thinking Inside the Box: Using Appearance Models and Context Based on Room Geometry (2010, methodology): “We start by detecting vanishing points corresponding to 3 orthogonal directions of the world using the vanishing point method from [1].”
Towards Holistic Scene Understanding: Feedback Enabled Cascaded Classification Models (2010, methodology): “For geometric labeling, we use the Indoor Layout Data [69] and assign each pixel to one of three geometry classes: ground, wall and ceiling.”
Efficient Structured Prediction with Latent Variables for General Graphical Models (2012, methodology): “Following existing approaches (Hedau et al., 2009; Lee et al., 2010), we employ F = 55 features based on geometric context (GC) and orientation maps (OM) and refer the interested reader to (Hoiem et al., 2007) and (Lee et al., 2009) for respective details.”
Generic 3D Representation via Pose Estimation and Matching (2016, methodology): “Table 4 provides the results of layout estimation using a simple NN classiﬁer on our representation along with two supervised baselines, showing that our representation (with no ﬁne-tuning) achieved a performance close to Hedau et al.’s [27] supervised method on this novel task.”
Factoring Shape, Pose, and Layout from the 2D Image of a 3D Scene (2017, background): “Past approaches to this [15, 27] have primarily posed this in terms of fitting a vanishing-point-aligned 3D box, which intrinsically cannot generalize to non-box environments.”
Understanding Indoor Scenes Using 3D Geometric Phrases (2013, methodology): “In practice, rather than learning the parameters for the feature vectors of the observation model, we use the confidence values given by SPM [15] for scene classification, from [12] for layout estimation, and from the DPM [8] for object detection.”
Scene Parsing by Integrating Function, Geometry and Appearance Models (2013, methodology): “2, we also test our algorithm on the UCB dataset and the UIUC dataset together with five state-of-the-art algorithms: Hedau 2009 [12], Wang 2010 [21], Lee 2010 [16], Del Pero 2011 [4] and Del Pero 2012 [3].”
What's Behind the Couch? Directed Ray Distance Functions (DRDF) for 3D Scene Reconstruction (2021, methodology): “Early work often used vanishing-point-aligned box [12, 23] trained on annotated data.”