Object Detection Using the Statistics of Parts

By Henry Schneiderman, Takeo Kanade. International Journal of Computer Vision, 2004

420 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/36cd88ed2c17a596001e9c7d89533ac46c28dec0

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

Citations in the corpus

Histograms of oriented gradients for human detection (2005, neutral): “There is an extensive literature on object detection, but here we mention just a few relevant papers on human detection [18,17,22,16,20].”
Non-maximum Suppression for Object Detection by Passing Messages Between Windows (2014, methodology): “Recently, a more common approach has been to adopt a greedy procedure [2, 3, 11] where the top-scoring window is declared an object, then neighboring windows are removed based on a hand-tuned threshold of a manually-designed similarity (distance between centers when the size ratio is within 0.”
Non-commercial Research and Educational Use including without Limitation Use in Instruction at Your Institution, Sending It to Specific Colleagues That You Know, and Providing a Copy to Your Institution's Administrator. All Other Uses, Reproduction and Distribution, including without Limitation Comm (2005, neutral): “…S kl is employed, with S kl defined as averages over reorientational motion of the second-order Legendre polynomials: S kl ¼ 1 2 < 3 cos h Z;k cos h Z;l À d kl > ð2Þ where h Z,k is the angle between the molecule-fixed k-axis and the space-fixed Z-axis which lies along the liquid-crystal director.”
FloatBoost learning and statistical face detection (2004, methodology): “The BDF classifier is reported to achieve result which compares favorably against the state-of-the-art face detection algorithms, such as the Schneiderman-Kanade method.”
Shape Classification Using the Inner-Distance (2007, background): “The first category (e.g., [3], [19], [14], [37], [15], [46], etc.) builds part models from a set of sample images and usually with some prior knowledge such as the number of parts.”
PKLot - A robust dataset for parking lot classification (2015, methodology): “In this vein, there are several object detection algorithms that can be used (Schneiderman & Kanade, 2004; Viola & Jones, 2004).”
Object Detection with Grammar Models (2011, neutral): “[20] Henry Schneiderman and Takeo Kanade.”
Regularized Max Pooling for Image Categorization (2014, methodology): “To avoid the problem of rigid alignment, various works have proposed to model objects or scenes by parts in a deformable configuration [1, 8, 11, 20, 21, 24].”
Weak Hypotheses and Boosting for Generic Object Detection and Recognition (2004, methodology): “Schneiderman and Kanade [ 20 ] used Boosting to improve an already complex classifier.”
Recognition by association via learning per-exemplar distances (2008, background): “Stateof-the-art detectors exist for only a few object categories: cars [17], pedestrians [5], and faces [17, 22] (all compact objects, well modeled by a sliding rectangle).”
Finding People in Images and Videos (2006, background): “Local “parts”-based detectors are also widely used in object and human recognition systems [Forsyth and Fleck 1997, Ioffe and Forsyth 1999, Schneiderman and Kanade 2000, Ronfard et al. 2002, Ramanan and Forsyth 2003, Sigal et al. 2003, Schneiderman and Kanade 2004].”
Peekaboom: a game for locating objects in images (2006, methodology): “[4,5,9,10]) rely on machine learning: train an algorithm to perform a visual task by showing it example images in which the task has already been performed.”
Robust object detection via soft cascade (2005, background): “It is often used implicitly by preﬁltering obvious non-faces based on various heuristic criteria, such as the presence of skin tone [7] or intensity variance [10].”
Generic object recognition with boosting (2006, methodology): “Additionally, Schneiderman and Kanade undertake rather specific object recognition as they train each object from different viewpoints.”
Detection and Localization of Multiple Objects (2006, background): “However, most parts-based recognition algorithms still suffer from very high computational costs, even when a sparse object representation is used [9] [10].”
Labeling images with a computer game (2004, methodology): “[17], for instance, introduced a method to locate human faces in still photographs.”
A discriminatively trained, multiscale, deformable part model (2008, background): “The notion that objects can be modeled by parts in a deformable configuration provides an elegant framework for representing object categories [1–3, 6,10, 12, 13,15, 16, 22].”
A Survey of Recent Advances in Face Detection (2010, methodology): “Schneiderman and Kanade [83] used wavelet variables to represent parts of the faces, which do not necessarily corresponds to semantic components.”