Unsupervised Learning of Models for Recognition

By Markus Weber, Max Welling, Pietro Perona. 2000

778 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

Unsupervised Learning of Visual Representations by Solving Jigsaw Puzzles (2016, methodology): “For instance [11,35] introduced methods to build a probabilistic representation of objects as constellations of parts.”
Unsupervised Visual Representation Learning by Context Prediction (2015, methodology): “astive divergence [20], deep Boltzmann machines [40], and Bayesian nonparametric models [43] use sampling to perform approximate inference. Others strengthen the independence assumptions in the model [48]. More recently, variational Bayesian methods for approximate inference have been proposed [25, ”
Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories (2004, methodology): “…et al. and against the maximum likelihood method of Fergus et al. We wish to emphasize at this point that previous work on object categories has been tested for the most part on 1 or 2 categories [3] with the exception of Weber et al [5] who tested on four and Fergus et al [8] who tested on six.”
Unsupervised Learning of Visual Representations Using Videos (2015, background): “f the work in the area of unsupervised learning can be broadly divided into three categories. The ﬁrst class of unsupervised learning algorithms focus on learning generative models with strong priors [17,49,42]. These al1http://www.aoa.org/patients-and-public/good-vision-throughoutlife/childrens-vis”
Discovering objects and their location in images (2005, background): “This may range from specifying the object’s location and segmentation, as in face detection [17, 23], to providing only auxiliary data indicating the object’s identity [1, 2, 8, 24].”
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.”
Semi-Supervised Self-Training of Object Detection Models (2005, neutral): “The work which is most similar to our work is that described in [24].”
Improving Generalization via Scalable Neighborhood Component Analysis (2018, background): “Most of the research in this area is focused on designing better invariant category representations (e.g. bag-of-words [45], pictorial models [5]).”
An expectation maximization approach to the synergy between image segmentation and object categorization (2005, methodology): “For faces the MRF-based detection system and the morphable model were trained using 100 hand-labelled face images from the database in [15], and the system was evaluated on the data set of [29].”
Some Objects Are More Equal Than Others: Measuring and Predicting Importance (2008, background): “We now have algorithms that can recognize individual objects accurately and quickly [1], classify scenes [2], and learn new categories with little supervision [ 3 ,4,5,6,7].”
An Affine Invariant Salient Region Detector (2004, background): “The selection of a set of image regions forms the first step in many computer vision algorithms, for example for computing image correspondences [2,17,19,20, 22], or for learning object categories [1,3,4,23].”
A Hierarchical Model of Shape and Appearance for Human Action Classification (2007, neutral): “Constellation models offer such a solution [10, 21].”
On Space-Time Interest Points (2005, neutral): “Thus clustering of spatio-temporal neighborhoods is similar to the idea of textons (Malik et al., 1999) used to describe image texture as well as to detect object parts for spatial recognition (Weber et al., 2000).”
Hide-and-Seek: Forcing a Network to be Meticulous for Weakly-Supervised Object and Action Localization (2017, methodology): “Weakly-supervised approaches have been proposed for various visual classification and localization tasks including object detection [55, 13, 9, 41, 3, 50, 43, 8, 32, 61, 40], semantic segmentation [33, 26] and visual attribute localization [2, 53, 57, 52, 39].”
Measuring and Predicting Importance of Objects in Our Visual World (2007, background): “A full description of complex scenes, such as the one shown in Fig.”
Object perception as Bayesian inference. (2004, background): “The mechanisms of bootstrapped learning are not well-understood although there has been some computer vision work (Weber et al 2000), see Section 4.”
Object Detection with Discriminatively Trained Part Based Models (2010, methodology): “One popular approach is to restrict part locations to a small discrete set of locations returned by an interest point detector [8], [9], [11].”
Object recognition with features inspired by visual cortex (2005, methodology): “we also tested our approach on a subset of the 101-object dataset using the exact same split as in [4] (the results are reported in Table 2) and an additional leaf database as in [24] for a total of five datasets that we refer to as the Caltech datasets in the following.”
Birdlets: Subordinate categorization using volumetric primitives and pose-normalized appearance (2011, background): “Constellation models [11] [48] and their computationally attractive variants [14] [27] are composed of a set of local part detectors together with one or more probability density functions describing the parts' relative locations.”