Top-down control of visual attention in object detection

By Aude Oliva, Antonio Torralba, Monica S. Castelhano, J. Henderson. Proceedings 2003 International Conference on Image Processing (Cat. No.03CH37429), 2003

553 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Cognitive Neuroscience, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/3cc6c4b2881c5607df9d3d6bb25ed94fd0add236

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 Cognitive Neuroscience

In Computer Vision and Pattern Recognition

Citations in the corpus

Visual attention detection in video sequences using spatiotemporal cues (2006, background): “[18] analyzed the global distributions of low-level features to detect the potential locations of target objects.”
State-of-the-Art in Visual Attention Modeling (2013, background): “[140], Torralba [92], and Bruce and Tsotsos [144] in that they are all based on the notion of selfinformation (local information).”
Visual search: a retrospective. (2011, unknown): “…scene types (Figures 4d and 4e; Castelhano & Heaven, 2010; Droll & Eckstein, 2008; Ehinger, Hidalgo-Sotelo, Torralba, & Oliva, 2009; Hidalgo-Sotelo, Oliva, & Torralba, 2005; Neider & Zelinsky, 2006; Oliva, Torralba, Castelhano, & Henderson, 2003; Torralba, Oliva, Castel-hano, & Henderson, 2006).”
Controllable Text-to-Image Generation (2019, background): “The attention mechanism has shown its efficiency in various research fields including image captioning [24, 30], machine translation [1], object detection [14, 29], and visual question answering [26].”
SUN: A Bayesian framework for saliency using natural statistics. (2008, methodology): “In most existing saliency algorithms, the features are calculated as responses of biologically plausible linear ﬁlters, such as DoG (difference of Gaussians) ﬁlters and Gabor ﬁlters (e.g., Itti & Koch, 2001; Itti et al., 1998; Oliva et al., 2003; Torralba et al., 2006).”
The natural/man-made distinction is made before basic-level distinctions in scene gist processing (2010, result): “First, the current results use human subject data to test a central assumption of an influential computational model of scene gist recognition, the Spatial Envelope model (Oliva & Torralba, 2001, 2006; Oliva et al., 2003; Torralba, 2003; Torralba et al., 2006), which has been essentially untested.”
Is bottom-up attention useful for object recognition? (2004, background): “It is conceivable, e.g., that in certain applications top-down knowledge can be very useful for visual processing in addition to the bottom-up saliency-based attention described here (see for instance [ 21 ]).”
Computational visual attention systems and their cognitive foundations: A survey (2010, methodology): “The contextual information is obtained from past search experiences in similar environments [Oliva et al. 2003; Torralba 2003b].”
Scene context guides eye movements during visual search. (2006, background): “Except for one recent demonstration of observers distributing their gaze on footpaths and doorways when looking for people ( Oliva, Torralba, Castelhano, & Henderson, 2003 ), the eVects of scene context on search remain largely unexplored.”
What do you see when you're surfing?: using eye tracking to predict salient regions of web pages (2009, background): “, about what is shown in the image and where to find specific objects) also play important roles in subsequent fixations [11, 22, 23].”
The Formation of Opportunity Beliefs: Overcoming Ignorance and Reducing Doubt (2007, background): “‘The gist benefi ts object detection mechanisms almost instantaneously as well as attention deployment and gaze control in cluttered scenes’ (Oliva et al., 2003; Torralba and Oliva, 2003).”
Human gaze control during real-world scene perception. (2003, methodology): “Scene identity can be apprehended and a scene schema retrieved very rapidly [47], and schema knowledge can then be used to limit initial fixations to scene areas likely to contain an object relevant to the current task [ 23 ,35].”
Region-Based Saliency Detection and Its Application in Object Recognition (2014, methodology): “One is top-down method, which is task-dependent and based on prior knowledge about the scenes [41], [51], [69], objects [61], face [40], and their interrelations [26].”
A Benchmark of Computational Models of Saliency to Predict Human Fixations (2012, background): “[29] provide evidence that top-down information dominates real-world image search processes, such that the influence of low-level salience information on search guidance is minimal and others show that context is important [71] [51].”
Static and space-time visual saliency detection by self-resemblance. (2009, methodology): “Most of the methods (Gao et al., 2008; Itti et al., 1998; Oliva et al., 2003; Zhang et al., 2009) based on Gabor or DoG filter responses require many design parameters such as the number of filters, type of filters, choice of the nonlinearities, and a proper normalization scheme.”
An Integrated Model of Top-Down and Bottom-Up Attention for Optimizing Detection Speed (2006, background): “Need to integrate top-down and bottom-up attentional influences: There have been few attempts to integrate both top-down and bottom-up attention [16].”
Visual memory for natural scenes: Evidence from change detection and visual search (2006, background): “For example, if one remembered the spatial position of a kitchen counter from the preview image, followed by a toaster target, one could bias search towards the likely location of the toaster on the counter (see Henderson, Weeks, & Hollingworth, 1999 ; Oliva, Torralba, Castelhano, & Henderson, 200”