Research Article

Learning to Model Task-Oriented Attention

Table 2

The table shows the average (Avg) and the corresponding standard deviations (STD) of the weight of attribute in each category. For every category, the bold weight is the first and the second is italic weight.

CategoryColorIntensityOrientationTargetCenter bias
AvgSTDAvgSTDAvgSTDAvgSTDAvgSTD

Airplane0.03190.00005−0.01540.000020.00980.000020.12010.00012−0.43440.00025
Bottle0.03460.000060.04240.000060.02940.000040.12060.00009−0.35860.00019
Car0.00730.000010.01120.00002−0.01590.000020.25750.00016−0.24180.00012
Chair0.02340.000030.05780.000060.10020.000110.27660.00013−0.13480.00008
Dog0.00660.000010.00750.000010.08480.000060.10650.00008−0.45560.00024
Horse0.02410.00004−0.00040.000000.02400.000030.14450.00011−0.31820.00031
Motorbike−0.00880.000020.01660.000020.02760.000040.20010.00015−0.27330.00025
Person−0.01310.00003−0.02910.000030.06380.000060.12410.00007−0.31590.00027