Visual features of intermediate complexity and their use in classification

By Shimon Ullman, Michel Vidal-Naquet, Erez Sali. Nature Neuroscience, 2002

677 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/d52be22dc0033293d335b6dc5cf3e3588c1fc0bc

How researchers see this paper

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

In Cognitive Neuroscience

Citations in the corpus

Building the gist of a scene: the role of global image features in recognition. (2006, background): “, openness, naturalness, expansion, symmetry) are implemented here as a weighted combination of global features, but spatial envelope properties could also be derived from other basis of low- or intermediate-level features (Ullman et al., 2002).”
50 Years of object recognition: Directions forward (2013, background): “They represent an attempt to compensate for the simplicity of localfeature-based representations, by using them to form dictionaries of more complex and compact object representations [195, 196, 179, 177, 176, 197, 178, 180, 198, 199, 100, 174, 200, 201, 202, 203].”
I2T: Image Parsing to Text Description (2010, background): “Most of component [66], [67], fragment [68], or part-based methods, the constellation models [69], and the pictorial models [70] belong to this process.”
A Dual-Route Approach to Orthographic Processing (2011, background): “…is thought to involve the development of letter combination detectors that best constrain word identity (the diagnosticity constraint), in the same way that parts of objects act as clues to object identity in certain theoretical approaches to visual object recognition (e.g., Ullman et al., 2002).”
Partially Distributed Representations of Objects and Faces in Ventral Temporal Cortex (2005, background): “Finally, these data help to separate the concept of representational types (object-form topography, Haxby et al., 2001; view-based, Riesenhuber & Poggio, 1999; Poggio & Edelman, 1990; structural, Biederman & Gerhardstein, 1993; intermediate complexity features, Ullman, Vidal-Naquet, & Sali, 2002 ) ”
On the Role of Object-Specific Features for Real World Object Recognition in Biological Vision (2002, result): “Those results are in good agreement with [10] where gray-value features of intermediate sizes where shown to have higher mutual information.”
Learning methods for generic object recognition with invariance to pose and lighting (2004, background): “…advocated the use of color, texture, and contours for image indexing applications [8], the detection of distinctive local features [20] [26] [25] [23], the use of global appearance templates [11] [10] [19], the extraction of silhouettes and edge information [14, 22, 8, 4, 19] and the use of…”
From simple innate biases to complex visual concepts (2012, background): “Formally, these are image features that have high mutual information with respect to images of a person (12).”
A Comparative Study of Local Matching Approach for Face Recognition (2007, background): “Recently, local matching approaches have shown promising results not only in face recognition [2], [15], [16], [20], [29], [33], [42], [47], [53] but also in other visual recognition tasks [45].”
Multiclass Object Recognition with Sparse, Localized Features (2006, background): “Among these surviving features, the 4x4 size dominates (figure 6), suggesting that this size yields the most informative features for this task [24].”
Bringing Semantics into Focus Using Visual Abstraction (2013, background): “(2) In the case that we want to condition upon two variables, we compute the CMI for each variable individually and take the minimum value [34].”
Semantic Classification of High-Resolution Remote-Sensing Images Based on Mid-level Features (2016, background): “It has been pointed out that mid-level features are more informative for object recognition [17].”
Cross-generalization: learning novel classes from a single example by feature replacement (2005, background): “Assume that for several object classes (e.g. horses and cows), a sufficient number of examples has been available to perform feature extraction by one of the existing methods [1] [17] [19].”