Your Local GAN: Designing Two Dimensional Local Attention Mechanisms for Generative Models
This work addresses the challenge of enhancing image generation quality in GANs for computer vision applications, representing an incremental improvement with novel attention mechanisms.
The paper tackles the problem of improving generative adversarial networks (GANs) by designing a new local sparse attention layer that preserves two-dimensional geometry, resulting in significant improvements such as FID score enhancement from 18.65 to 15.94 on ImageNet and better visual quality.
We introduce a new local sparse attention layer that preserves two-dimensional geometry and locality. We show that by just replacing the dense attention layer of SAGAN with our construction, we obtain very significant FID, Inception score and pure visual improvements. FID score is improved from $18.65$ to $15.94$ on ImageNet, keeping all other parameters the same. The sparse attention patterns that we propose for our new layer are designed using a novel information theoretic criterion that uses information flow graphs. We also present a novel way to invert Generative Adversarial Networks with attention. Our method extracts from the attention layer of the discriminator a saliency map, which we use to construct a new loss function for the inversion. This allows us to visualize the newly introduced attention heads and show that they indeed capture interesting aspects of two-dimensional geometry of real images.