SalGAN: visual saliency prediction with generative adversarial networks
Pan, Junting and Sayrol, Elisa and Giro-i-Nieto, Xavier and Canton Ferrer, Cristian and Torres, Jordi and McGuinness, Kevin and O'Connor, Noel E. (2017) SalGAN: visual saliency prediction with generative adversarial networks. In: CVPR SUNw: Scene Understanding Workshop 2017, July 26, 2017, Hawaii.
Full text available as:
We introduce SalGAN, a deep convolutional neural network for visual saliency prediction trained with adversarial examples. The first stage of the network consists of a generator model whose weights are learned by back-propagation computed from a binary cross entropy (BCE) loss over downsampled versions of the saliency maps. The resulting prediction is processed by a discriminator network trained to solve a binary classification task between the saliency maps generated by the generative stage and the ground truth ones. Our experiments show how adversarial training allows reaching state-of-the-art performance across different metrics when combined with a widely-used loss function like BCE. Our results can be reproduced with the source code and trained models available at https://imatge-upc.github. io/saliency-salgan-2017/.
Archive Staff Only: edit this record