Robust and compact video descriptor learned by deep neural network | IEEE Conference Publication | IEEE Xplore