Home People Research Publications Demos
News Jobs Prospective
About Internal

Semantic Fisher Vectors for Scene Classification
(MIT SUN Dataset)

Scene images were represented using Semantic Fisher vectors and classified using a linear SVM. Examples of correct classification are shown below. The decision values were converted into probability vectors using scaling and softmax. Bar charts depict the probabilities of the top 30 SUN classes, given the images. Labels for top 5 classes are also provided.
ImagesSoftmax of SVM scores