Neural Knitworks: Patched Neural Implicit Representation Networks

Abstract

Coordinate-based Multilayer Perceptron (MLP) networks, despite being capable of learning neural implicit representations, are not performant for internal image synthesis applications. Convolutional Neural Networks (CNNs) are typically used instead for a variety of internal generative tasks, at the cost of a larger model. We propose Neural Knitwork, an architecture for neural implicit representation learning of natural images that achieves image synthesis by optimizing the distribution of image patches in an adversarial manner and by enforcing consistency between the patch predictions. To the best of our knowledge, this is the first implementation of a coordinate-based MLP tailored for synthesis tasks such as image inpainting, super-resolution, and denoising. We demonstrate the utility of the proposed technique by training on these three tasks. The results show that modeling natural images using patches, rather than pixels, produces results of higher fidelity. The resulting model requires 80% fewer parameters than alternative CNN-based solutions while achieving comparable performance and training time.

Published in Pattern Recognition

References in corpus (10)

• Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains
• Deep Image Prior
• NeRF++: Analyzing and Improving Neural Radiance Fields
• Blind Super-Resolution Kernel Estimation using an Internal-GAN
• On the Spectral Bias of Neural Networks
• Swapping Autoencoder for Deep Image Manipulation
• Structure-Preserving Image Super-Resolution
• Super-Resolution via Deep Learning
• Exploring Differential Geometry in Neural Implicits
• Learning Neural Acoustic Fields

Cited by in corpus (2)

• Neural Fields in Visual Computing and Beyond
• Compositional Meta-Learning for Mitigating Task Heterogeneity in Physics-Informed Neural Networks