Fast-AT is an automatic thumbnail generation system based on deep neuralnetworks. It is a fully-convolutional deep neural network, which learnsspecific filters for thumbnails of different sizes and aspect ratios. Duringinference, the appropriate filter is selected depending on the dimensions ofthe target thumbnail. Unlike most previous work, Fast-AT does not utilizesaliency but addresses the problem directly. In addition, it eliminates theneed to conduct region search on the saliency map. The model generalizes tothumbnails of different sizes including those with extreme aspect ratios andcan generate thumbnails in real time. A data set of more than 70,000 thumbnailannotations was collected to train Fast-AT. We show competitive results incomparison to existing techniques.
展开▼