SalGAN: Visual Saliency Prediction with Generative Adversarial Networks
This work addresses the problem of accurately predicting visual saliency for applications in computer vision, representing an incremental improvement over existing methods.
The authors tackled visual saliency prediction by introducing SalGAN, a deep convolutional neural network trained with adversarial examples, which achieved state-of-the-art performance across different metrics when combined with binary cross entropy loss.
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/.