GeoPos: A Minimal Positional Encoding for Enhanced Fine-Grained Details in Image Synthesis Using Convolutional Neural Networks
This addresses a fundamental shortcoming in image generation affecting all models, potentially improving fine-grained details for applications requiring high-fidelity synthesis.
The paper tackles the persistent problem of image generative models failing to recreate intricate geometric features like human hands by augmenting convolution layers with a single input channel of relative n-dimensional Cartesian coordinates, showing this drastically improves image quality across Diffusion Models, GANs, and VAEs.
The enduring inability of image generative models to recreate intricate geometric features, such as those present in human hands and fingers has been an ongoing problem in image generation for nearly a decade. While strides have been made by increasing model sizes and diversifying training datasets, this issue remains prevalent across all models, from denoising diffusion models to Generative Adversarial Networks (GAN), pointing to a fundamental shortcoming in the underlying architectures. In this paper, we demonstrate how this problem can be mitigated by augmenting convolution layers geometric capabilities through providing them with a single input channel incorporating the relative n-dimensional Cartesian coordinate system. We show this drastically improves quality of images generated by Diffusion Models, GANs, and Variational AutoEncoders (VAE).