An Integrated Model of Top-Down and Bottom-Up Attention for Optimizing Detection Speed

By Vidhya Navalpakkam, Laurent Itti. Computer Vision and Pattern Recognition, 2006

423 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/b92cff15dfc13dc83a253c2212ce624628941bef

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

In Computer Vision and Pattern Recognition

Citations in the corpus

A Novel Multiresolution Spatiotemporal Saliency Detection Model and Its Applications in Image and Video Compression (2010, background): “We believe that the potential of our work lies in many other engineering ﬁelds besides image and video compression, such as image retrieval and object recognition [13], [54], [55].”
Learning to Detect A Salient Object (2007, background): “Recent studies [18, 22, 26] demonstrated that visual attention helps object recognition, tracking, and detection as well.”
Attentive Systems: A Survey (2017, background): “Navalpakkam and Itti (2006) propose a model that combines both bottom-up as well as top-down attentional influences.”
Exploiting local and global patch rarities for saliency detection (2012, background): “The human map is usually smoothed by convolving with a small 7Available at: http://mmas.comp.nus.edu.sg/NUSEF.html 8 AIM: http://www-sop.inria.fr/members/Neil.Bruce/ GBVS: http://www.klab.caltech.edu/˜harel/ SRM & ICL: http://www.klab.caltech.edu/˜xhou/ Itti & Surprise: http://ilab.usc.edu/toolkit/ ”
Attention, please! A survey of neural attention models in deep learning (2021, background): “Navalpakkam and Itti introduce a derivative of NVT which can deal with top-down cues (Navalpakkam and Itti 2006).”
Few-Shot Semantic Segmentation with Prototype Learning (2018, neutral): “FG-BG has played an important role in the pre-processing step for object detection [24], face detection [21] and motion detection [25].”
Deep Contrast Learning for Salient Object Detection (2016, background): “Though early work primarily focused on predicting eye-fixations in images, research has shown that salient object detection, which emphasizes object-level integrity of saliency prediction results, is more useful and can serve as a pre-processing step for a variety of computer vision and image proces”
What Do Different Evaluation Metrics Tell Us About Saliency Models? (2016, methodology): “Consider a detection application of saliency such as object and motion detection, surveillance, localization and mapping, and segmentation [1], [17], [27], [28], [43], [50], [59], [91].”
Quantitative Analysis of Human-Model Agreement in Visual Saliency Modeling: A Comparative Study (2013, methodology): “Context [3], [35], [45], gain modulation of features for target detection [4], [5], [80], and use of target detectors tuned to specific objects [50], [69] has been used for modeling topdown attentional effects.”
Salient Object Detection via Structured Matrix Decomposition (2017, background): “It has received substantial attention over the last decade due to its wide range of applications in computer vision, such as object detection and recognition [1]–[4], content-based image retrieval [5], [6] and context-aware image resizing [7]–[10].”
Computational visual attention systems and their cognitive foundations: A survey (2010, methodology): “More elaborated is the tuning of the conspicuity maps [Milanese et al. 1994; Hamker 2005], but biologically most plausible and also most useful from an engineering perspective is the approach to already bias the feature types [Tsotsos et al. 1995; Frintrop et al. 2005; Navalpakkam and Itti 2006a].”
Saliency Detection via Absorbing Markov Chain (2013, neutral): “It has been effectively applied to numerous computer vision tasks such as content based image retrieval [32] , image segmentation [30], object recognition [24] and image adaptation [21].”
Attentional Pooling for Action Recognition (2017, methodology): “3.1, our inspiration for combining class-speciﬁc and class-agnostic classiﬁers (i.e. top-down and bottom-up attention respectively), came from the Neuroscience literature on integrating top-down and bottom-up attention [31].”