Intrinsic Scene Properties from a Single RGB-D Image

By Jonathan T. Barron, Jitendra Malik. 2013 IEEE Conference on Computer Vision and Pattern Recognition, 2013

306 citations recorded by the source corpus.

Subfields: Computational Mechanics, Computer Graphics and Computer-Aided Design, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/74a08647f3852750b3915567d4f9f76d2f79f959

How researchers see this paper

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

In Computational Mechanics

In Computer Graphics and Computer-Aided Design

In Computer Vision and Pattern Recognition

Citations in the corpus

Learning to predict indoor illumination from a single image (2017, methodology): “Because these properties are intrinsically tied to each other, estimating each of them often relies on reconstructing the others too [Barron and Malik 2013b].”
Indoor Scene Understanding with RGB-D Images: Bottom-up Segmentation, Object Detection and Semantic Segmentation (2015, background): “For example, recent works have considered 3D reconstruction tasks such as real-time scene reconstruction [24], and recovering high fidelity albedo, shape and illumination [4].”
Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-scale Convolutional Architecture (2014, methodology): “[21], Barron and Malik [3, 2], and most recently by Wang et al.”
CGIntrinsics: Better Intrinsic Image Decomposition through Physically-Based Rendering (2018, methodology): “Several recent methods seek to improve decomposition quality by integrating surface normals or depths from RGB-D cameras [18,19,20] into the optimization process.”
DeepLiDAR: Deep Surface Normal Guided Depth Prediction for Outdoor Scene From Sparse LiDAR Data and Single Color Image (2018, methodology): “Traditional methods use color as the guidance to compute local affinity or discontinuity [18, 14, 44, 2, 12, 54, 58, 1].”
Layered RGBD scene flow estimation (2015, background): “[6] leverage the depth for intrinsic image decomposition of natural scenes.”
FPM: Fine Pose Parts-Based Model with 3D CAD Models (2014, background): “The recent availability of affordable depth sensors has also led to an increase in the use of RGB-D data [21,11,3,30,26] to extract various types of 3D information from scenes such as depth, object pose, and intrinsic images.”
OpenRooms: An Open Framework for Photorealistic Indoor Scene Datasets (2020, methodology): “Barron13 [8] Gardner17”
Dense planar SLAM (2014, methodology): “Depth maps are further corrupted by the noise model proposed by Barron and Malik [4] to create data closer to the Kinect sensor.”
OpenRooms: An End-to-End Open Framework for Photorealistic Indoor Scene Datasets (2020, methodology): “Barron13 [4] Gardner17 [17] Garon19 [18] Li20 [30] Ours Barron13 [4] 23.”
Inverse Rendering for Complex Indoor Scenes: Shape, Spatially-Varying Lighting and SVBRDF From a Single Image (2019, methodology): “Our spherical Gaussians representation for local lighting is demonstrably better than spherical harmonics in Barron13 [4], Sengupta19 [44] and Garon19 [18].”
SceneNN: A Scene Meshes Dataset with aNNotations (2016, methodology): “Qualitatively, the reflectance by Chen and Koltun [4] and the shading by Barron and Malik [2] closely resembles the reflectance and shading observed in the input images.”
NeRD: Neural Reflectance Decomposition from Image Collections (2020, background): “Intrinsic images describes the task of separating a scene into its underlying physical constituents, such as reflectance and shading [5–7, 43, 62].”
Neural RGB-D Surface Reconstruction (2021, methodology): “We apply noise and artifacts, similar to those of a real depth sensor [1, 3, 30, 31].”
Estimating the 3D Layout of Indoor Scenes and Its Clutter from Depth Sensors (2013, background): “Additionally, the superiority of RGB-D sensors when compared to more traditional imagery has been demonstrated for the tasks of semantic segmentation [25, 26, 8], inferring support relations [26], 3D detection [14] or estimating physical properties of images [2].”
DeepContour: A deep convolutional feature learned by positive-sharing loss for contour detection (2015, background): “Accurately detecting object contours serves as the basis for many tasks, such as image segmentation [2], scene understanding [3] and object detection [17, 30, 48].”