Learning hierarchical models of scenes, objects, and parts

By Erik B. Sudderth, Antonio Torralba, William T. Freeman, Alan S. Willsky. 2005

380 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Probabilistic & Bayesian Methods

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

In Probabilistic & Bayesian Methods

Citations in the corpus

LabelMe: A Database and Web-Based Tool for Image Annotation (2008, neutral): “Also, we show how to extend the dataset to automatically enhance object labels with WordNet, discover bject parts, recover a depth ordering of objects in a scene, and increase the number of labels using minimal user supervision and images from the web.”
Using Multiple Segmentations to Discover Objects and their Extent in Image Collections (2006, background): “Therefore, what we need is a way to group visual words spatially [8, 24] to make them more descriptive.”
Context as Supervisory Signal: Discovering Objects with Predictable Context (2014, background): “The problem of generative image modeling has a long history in computer vision [1,4,25,30,31,32], but historically these algorithms have performed poorly for object recognition problems, especially compared to the discriminative methods that have, of late, largely displaced them in the field.”
Sharing Visual Features for Multiclass and Multiview Object Detection (2007, background): “Transferring knowledge between objects to improve generalization has also been studied in several recent papers [4], [17], [34].”
What, where and who? Classifying events by scene and object recognition (2007, methodology): “Mathematically, our paper is closest in spirit with Sudderth et al [18].”
Moving vistas: Exploiting motion for describing scenes (2010, background): “The large intra-class variation due to unconstrained settings and the complex underlying physics present challenging problems in modeling scene dynamics.”
Describing the scene as a whole: Joint object detection, scene classification and semantic segmentation (2012, background): “Sudderth et al [23] relate scenes object and parts, but not segmentation, in a generative model.”
Learning Object Categories From Internet Image Searches (2010, methodology): “Sudderth et al. [38], [39] construct hierarchical sceneobject-part models based on LDA.”
Reinforcement Learning for Visual Object Detection (2016, background): “This includes sharing computation among neighboring image regions[40] [35] or among the different classifiers for multi-class detection problems[20].”
Discriminative Models for Multi-Class Object Layout (2009, background): “There has been a wide variety of work in the last few years on contextual modeling in image parsing (Torralba et al. 2004; Sudderth et al. 2005; Hoiem et al. 2008; Galleguillos et al. 2008; Shotton et al. 2006; He et al. 2004; Anguelov et al. 2005).”
Animals on the Web (2006, methodology): “We compute a likelihood for each image according to its associated word vector and the word likelihoods found by LDA.”
A Linear Time Histogram Metric for Improved SIFT Matching (2008, methodology): “The SIFT descriptor has proven to be successful in applications such as object recognition [3, 7–9], object class detection [10–12], image retrieval [13–16], robot localization [17], building panoramas [18] and image classification [19].”
MedLDA: maximum margin supervised topic models (2012, background): “Examples of USTM 3 include DiscLDA and the scene understanding models (Sudderth et al., 2005; Li et al., 2009), whereas sLDA is an example of DSTM.”
Semantic Label Sharing for Learning with Many Categories (2010, background): “Other approaches transfer information across object categories by sharing a common set of parts [16, 17], by sharing transformations across different instances [18–20], or by sharing a set of prototypes [21].”
Bottom-Up Segmentation for Top-Down Detection (2013, background): “Over the past few years, we have witnessed a push towards holistic approaches that try to solve multiple recognition tasks jointly [29, 6, 20, 18, 33].”
VisKE: Visual knowledge extraction and question answering by visual verification of relation phrases (2015, background): “, ‘IsA(horse, animal)’, ‘PartOf(wheel, car)’, ‘HasA(horse, leg)’) are easier to acquire and have been previously explored in computer vision [35, 25, 6], more complex actioncentric relationships (e.”
Transfer learning for image classification with sparse prototype representations (2008, background): “Transfer learning for visual category recognition has received significant attention in recent years [8, 18, 12, 24, 9, 10, 20].”
OPTIMOL: Automatic Online Picture Collection via Incremental Model Learning (2007, background): “A number of systems based on the bag of words model representation have shown to be effective for object and scene recognition (Fergus et al. 2005b; Sivic et al. 2005; Fei-Fei and Perona 2005; Sudderth et al. 2005b; Bosch et al. 2006 ;C surka et al.2004; Sivic and Zisserman 2003).”