Towards Deeper Generative Architectures for GANs using Dense connections
This is an incremental improvement for generative modeling researchers, focusing on architectural tweaks to enhance GAN performance.
The paper tackles the problem of improving image generation quality in GANs by incorporating skip connections and dense layers from image classification into Fisher GANs, finding that these modifications produce better images than the baseline with only slight effects from the number of connections.
In this paper, we present the result of adopting skip connections and dense layers, previously used in image classification tasks, in the Fisher GAN implementation. We have experimented with different numbers of layers and inserting these connections in different sections of the network. Our findings suggests that networks implemented with the connections produce better images than the baseline, and the number of connections added has only slight effect on the result.