What are the attributes of a good computer?

Computer Vision: How attributes are assigned/extracted from Images?

• I am interested to know about how we assign attributes to images, like natural, smiling etc. on the coding level. I know you can use GIST for extracting attributes, but I am looking for a better explanation.

• Answer:

  One of the ways to tag images with attributes is to train a set of classifiers for individual visual attributes whose list has been manually specified. [Su and Jurie's approach https://jurie.users.greyc.fr/papers/12_SEMATR_IJCV.pdf In their work, six groups of semantic attributes are introduced to cover the spectrum of (1) global scenes (e.g. train station, bedroom), (2) local scene elements (e.g. sky, tree), (3) color (e.g. green, red), (4) shape (e.g. box, cylinder), (5) material (e.g. leather, wood) and (6) object (e.g. face, motorbike). It makes a total of 110 different attributes. This is similar to the idea presented by Li et al. in their Object bank paper, in which an image is represented as the localized outputs of object detectors. In these methods, classifiers are trained for each individual semantic attribute and the classifier outputs are used to represent images. http://vision.stanford.edu/projects/objectbank/