A principled approach to detecting surprising events in video

By Laurent Itti, Pierre Baldi. Computer Vision and Pattern Recognition, 2005

435 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/71860dd7a275f7794f2054829062849a3ea4a73f

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

Saliency and Gist Features for Target Detection in Satellite Images (2011, background): “Surprise Channel: We recently proposed an enhanced saliency model, which exploits a new Bayesian deﬁnition of surprise to predict human perceptual salience in space and time [55]–[57].”
SUN: A Bayesian framework for saliency using natural statistics. (2008, methodology): “In (Itti & Baldi, 2005, 2006), the random locations are drawn from a uniform spatial distribution over each image frame.”
Key-segments for video object segmentation (2011, background): “, [12, 9]), recent work suggests that higher-level saliency may actually be learned from labeled data of segmented objects [19, 1, 6, 8], drawing on classic Gestalt cues.”
Visual adaptation and novelty responses in the superior colliculus (2011, methodology): “This can be achieved through incremental updating over time of a Bayesian prior, which can then bias the processing of incoming sensory data (Itti & Baldi, 2005; Stocker & Simoncelli, 2006).”
Bayesian surprise attracts human attention (2005, background): “This work extends Itti and Baldi (2006), through a more complete exposition of the theory and of the new proposed unit of surprise (the ‘‘wow”), simple examples of how surprise may be computed, and a broader set of experiments and comparisons with competing theories and models.”
Static and space-time visual saliency detection by self-resemblance. (2009, methodology): “PERFORMANCE IN PREDICTING HUMAN EYE FIXATIONS WHEN VIEWING VIDEOS [15].”
Novelty or Surprise? (2013, methodology): “Itti and Baldi (2005, 2006, 2009) argue that this deﬁnition has key advantages over alternatives in being more principled, more widely applicable, and more able to account for what attracts human visual attention.”
Unmasking the Abnormal Events in Video (2017, background): “s running outside from a bank (abnormal event) is another example. Although what is considered abnormal depends on the context, we can generally agree that abnormal events should be unexpected events [10] that occur less often than familiar (normal) events. As it is generally impossible to ﬁnd a suf”
Video anomaly detection based on local statistical aggregates (2012, background): “Itti and Baldi consider low-level feature descriptors at every location [10] and use possion statistics for modeling nominal activity.”
Clustering of Gaze During Dynamic Scene Viewing is Predicted by Motion (2011, background): “The study of luminance, color, and their respective retinotopic gradients (contrast) as a predictor or major contributor of the human visual salience map is well documented [13, 21, 76] and typically employed in computational models of saliency [71, 77 ].”
Towards Universal Object Detection by Domain Attention (2019, background): “First, feature-based attention is well known to be used in mammalian vision as a mechanism to adapt perception to different tasks and environments [61, 37, 52, 18, 60].”
Abnormal event detection in crowded scenes using sparse representation (2013, methodology): “There are also saliency feature [14] and graph-based non-linear dimensionality reduction method [30].”
Sparse reconstruction cost for abnormal event detection (2011, methodology): “The saliency features are extracted and associated by graph model in [12].”
Visual Place Recognition: A Survey (2016, methodology): “larities with the problem of change-point detection in video segmentation [49], [50], and used change-point detection algorithms such as Bayesian surprise [50] and segmented regression”
Learning to Detect A Salient Object (2007, methodology): “Most existing visual attention approaches are based on the bottom-up computational framework [3, 6, 8, 9, 10, 11, 19, 25] because visual attention is in general unconsciously driven by low-level stimulus in the scene such as intensity, contrast, and motion.”
Spatiotemporal Saliency in Dynamic Scenes (2010, background): “[21], and the “surprise” model proposed by Itti and Baldi [17], [18].”
Of Bits and Wows: A Bayesian Theory of Surprise with Applications to Attention (2009, methodology): “The application presented here extends previous similar experimentations (Itti & Baldi, 2005, 2006, 2009) by adding a new dataset, comparing two different model classes, Gaussian data with a Gaussian prior, and Poisson data with a Gamma prior, and introducing a new temporal analysis which…”