Results 1 - 10 of 37,002

Table 1. The results of evaluation for image adaptation based on manual attention modeling.

in Image adaptation based on attention model for small-form factor devices
by Li-qun Chen, Xing Xie, Wei-ying Ma, Hong-jiang Zhang, He-qin Zhou 2003
Cited by 4

Table 1. Descriptive statistics for attention-enhanced and standard backpropagation performance.

in unknown title
by unknown authors
"... In PAGE 1: ... We ran 100 different simulations in which both networks used the same random initial weight configuration, but for two of the then initial weight configurations, the standard backpropagation network did not converge, while the attention-enhanced network did. Figure 1A and Table1... In PAGE 2: ...30 175.65 Range 1278 531 Minimum 975 279 Maximum 2253 810 Table1 . Descriptive statistics for attention-enhanced and standard backpropagation performance.... ..."

Table 1 represents the correlation coefficients, for the six considered images, of the computational saliency map on one hand and three human attention maps on the other hand. The three human attention maps are: the mean human attention map (first row), the individual human attention map with the highest correlation coefficient (second row), and the individual human attention map with the lowest correlation coefficient (third row). For most of the images the correlation between the computational saliency and the mean human attention map is quite high. The two landscape images (swissalps and forest), where the stimuli are more bottom- up driven, give rise to particularly highly correlated human and computational maps of attention. For the images containing traffic signs, where more top-down information is present, the computational and the human attentional behavior are less correlated. Figure 5 visually illustrates the correlation between the computational map and human maps (mean and individuals) for the image swissalps .

in Electronic Letters on Computer Vision and Image Analysis 3(1):13-24, 2004 Empirical Validation of the Saliency-based Model of Visual Attention
by Nabil Ouerhani, Roman Von Wartburg, Heinz Hügli, René Müri 2003
"... In PAGE 8: ... Table1 : Correlation coefficients between the computational map of attention on one hand and mean and indi- vidual human maps of attention on the other hand. Some human subjects have better correlation with the computational saliency than the mean of all subjects.... ..."

TABLE 2. Initial network training

in unknown title
by unknown authors 2002
Cited by 1

Table-1: Visual Attention Values assigned to the 64 sub-image blocks. Imagine the grid to be superimposed on the given image shown in Figure-1.

in The packetvideo applications are being accessed by...
by Karun B. Shimoga

TABLE IV. MATCHING ATTENTIONAL NODES BETWEEN INDIVIDUAL IMAGE ACTIVATION SUMMING AND RECONSTRUCTED IMAGE N Number / % matching nodes in top N locations 12 5 / 42%

in Image Mapping and Visual Attention on a Sensory Ego-Sphere
by Katherine Achim Fleming, Richard Alan, Peters Ii, Robert E. Bodenheimer

Table 1. Model parameters for both attentional conditions. Name Symbol fully attended poorly attended Linear gainy

in Attentional Modulation of Human Pattern Discrimination Psychophysics Reproduced by a Quantitative Model
by Laurent Itti, Jochen Braun, Dale K. Lee, Christof Koch 1999
"... In PAGE 5: ... A multidimensional downhill simplex with simulated annealing overhead was used to minimize the root-mean-square distance between the quantitative predictions of the model and the human data [4]. The best- t parameters obtained independently for the \fully attended quot; and \poorly attended quot; conditions are reported in Table1 . The model apos;s simultaneous ts to our entire dataset are plotted in Figure 1 for both conditions.... In PAGE 6: ... Using poorly attended parameters, except for = 2:9 and = 2:1 (grey curves), yielded steep non-linear contrast response, and intermediary tuning (same width as NF). In Table1 , attention had the following signi cant e ects on the model apos;s param- eters: 1) Both pooling exponents ( ; ) were higher; 2) the tuning width ( ) was narrower; 3) the linear gain (A) and associated activity-independent inhibition (S) were lower; and 4) the background activity of the pooling stage was lower. This yielded increased competition between lters: The network behaved more like a winner-take-all under full attention, and more like a linear network of independent units under poor attention.... In PAGE 6: ... Although the new t was not as accurate as that obtained with all parameters allowed to vary, it appeared that a simple modi cation of the pooling exponents well captured the e ect of attention (Figure 1). Hence, the \poorly attended quot; parameters of Table1 well described the \poorly attended quot; data, and the same parameters except for = 2:9 and = 2:1 well described the \fully attended quot; data. A variety of other simple parameter modi cations were also tested, but none except for the pooling exponents ( ; ) could fully account for the attentional modulation.... ..."
Cited by 2

Table 1.3. The Image of the Artificial Network

in 1 Some Open Problem Sets for Generalized
by Patrick Doreian

Table 1 The image processing tasks categorised into a two-dimensional taxonomya

in Image Processing With Neural Networks - a Review
by Michael Egmont-Petersen, Dick de Ridder, Heinz Handels
"... In PAGE 3: .... Scene characterisation. A complete description of the scene possibly including lighting conditions, context, etc. Table1 contains the taxonomy of image processing algorithms that results from combining the steps of the image processing chain with the abstraction level of the input data. 3.... ..."

Table 1 The image processing tasks categorised into a two-dimensional taxonomya

in Image processing with neural networks -- a review
by M. Egmont-Petersen , D. De Ridder , H. Handels 2001
"... In PAGE 2: .... Scene characterisation. A complete description of the scene possibly including lighting conditions, context, etc. Table1 contains the taxonomy of image processing algorithms that results from combining the steps of the image processing chain with the abstraction level of the input data. 3.... ..."
Results 1 - 10 of 37,002