Stacked Hierarchical Labeling

By Daniel Munoz, J. Bagnell, Martial Hebert. 2010

249 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Natural Language Processing

Citations in the corpus

Convolutional Pose Machines (2016, methodology): “In contrast, methods based on a sequential prediction framework [29] learn an implicit spatial model with potentially complex interactions between variables by directly training an inference procedure, as in [22, 25, 31, 41].”
Designing deep networks for surface normal estimation (2014, background): “Additionally, both coarse and local networks treat every pixel independently; our fusion network can also be thought of applying a form of learned reasoning akin to [26, 35] on our outputs.”
Recurrent Convolutional Neural Networks for Scene Parsing (2013, background): “, 2010 [15] Gist, pyramid histogram of oriented gradients, color Histogram CIELab, relative relocation, hierarchical region representation.”
Reinforcement and Imitation Learning via Interactive No-Regret Learning (2014, background): “[1] Perhaps surprisingly, it has also become central in developing predictors for complex output spaces, e.g. sets and lists [22], parse trees [10], image parsing [20, 25] and natural language understanding[12].”
Learning message-passing inference machines for structured prediction (2011, background): “Similarly, from our point of view, [13] implements a ”half-pass” of hierarchical meanfield message passing by descending once down a hierarchy making contextual predictions.”
Pose Machines: Articulated Pose Estimation via Inference Machines (2014, methodology): “Inference machines have been previously studied in image and point cloud labeling applications [8,26].”
Scene parsing with Multiscale Feature Learning, Purity Trees, and Optimal Covers (2012, methodology): “A simple attention function a is used to mask the feature vector map with each component C k , producing a set of K masked feature vector patterns { F (cid:84) C k } , ∀ k ∈ { 1 , . . . , K } .”
Just-In-Time Learning for Fast and Flexible Inference (2014, background): “A related line of work is that of inference machines [14, 15, 17, 13].”
Advanced Structured Prediction (2014, methodology): “Another group of methods consider solving the Argmin problem via cascading (Felzenszwalb and McAllester, 2007; Weiss and Taskar, 2010; Munoz et al., 2010), where inference is run in stages from a coarse to fine level of abstraction.”
Fast LIDAR-based road detection using fully convolutional neural networks (2017, unknown): “LoDNN (our) 94.07 92.03 92.81 95.37 18 Up-Conv-Poly [21] 93.83 90.47 94.00 93.67 80 DDN [4] 93.43 89.67 95.09 91.82 2000 FTP [5] 91.61 90.96 91.04 92.20 280 FCN-LC [22] 90.79 85.83 90.87 90.72 30 HIM [23] 90.64 81.42 91.62 89.68 7000 NNP [24] 89.6886.50 89.67 5000 RES3D-Velo [9] 86.58 78.34 82.63 90”
Efficient deep models for monocular road segmentation (2016, methodology): “Global methods using Conditional Random Fields (CRF) [15], [19], Boosting Algorithms [18] and hierarchical approaches [13] were proposed to address this issue.”
Relating Things and Stuff via ObjectProperty Interactions (2014, methodology): “In Table 2a,(2) our ACRF model outperforms state-of-the-art methods [13], [21], [32], [44], [45] in the percentage of pixels correctly classified as either one of the eight background classes, or a general foreground class (referred to as global accuracy).”
Recurrent Convolutional Neural Networks for Scene Labeling (2014, background): “, 2010 [8] GIST, PYRAMID HISTOGRAM OF ORIENTED GRADIENTS, COLOR HISTOGRAM CIELAB, RELATIVE RELOCATION, HIERARCHICAL REGION REPRESENTATION.”
Relating Things and Stuff by High-Order Potential Modeling (2012, methodology): “We compare our full ACRF model with [1,20,21,12,13] on Stanford Background (refer as Stanford) dataset [1].”
Pylon Model for Semantic Segmentation (2011, methodology): “We note that our approach performs considerably better than state-of-the-art including the CRF-based method [10], the poolbased methods [11, 18], and the approach based on the same gPb-based tree [22].”
Computer Vision for Autonomous Vehicles: Problems, Datasets and State-of-the-Art (2017, background): “Munoz et al. (2010) propose an alternative to standard in-”
Instance Segmentation of Indoor Scenes Using a Coverage Loss (2014, methodology): “Like previous work in semantic segmentation [2] [8] [4] [9], we make use of hierarchical segmentations, referred to as segmentation trees, to limit the search space to a more manageable size.”
Deep hierarchical parsing for semantic segmentation (2015, methodology): “CRF based image models have been quite successful in jointly modeling the appearance and structure of an image; [7, 16, 15, 14] use CRFs to combine unary potentials obtained from the visual features of super-pixels with the neighborhood constraints.”
Multi-scale context for scene labeling via flexible segmentation graph (2016, methodology): “All the baselines are divided into two categories: (1) Modeling scene labeling using CRFs or MRFs, including TextonBoost (TB, [35]), relative location prior (RL, [37]), hierarchical CRF (HCRF, [42]), dynamic hybrid MRF (DHM, [23]), and full-connected CRF (FCRF, [59]); (2) Modeling scene labeling usi”