Image Generators with Conditionally-Independent Pixel Synthesis

TL;DR

This paper introduces a novel image generator architecture that computes each pixel independently based on a latent vector and pixel coordinate, eliminating the need for spatial convolutions, and achieves comparable quality to traditional methods.

Contribution

The paper proposes a new pixel-wise independent generator architecture that challenges the reliance on spatial convolutions in image synthesis models.

Findings

Achieves similar quality to state-of-the-art convolutional generators.

Exhibits unique properties due to its independent pixel computation.

Analyzes modeling capabilities of the new architecture.

Abstract

Existing image generator networks rely heavily on spatial convolutions and, optionally, self-attention blocks in order to gradually synthesize images in a coarse-to-fine manner. Here, we present a new architecture for image generators, where the color value at each pixel is computed independently given the value of a random latent vector and the coordinate of that pixel. No spatial convolutions or similar operations that propagate information across pixels are involved during the synthesis. We analyze the modeling capabilities of such generators when trained in an adversarial fashion, and observe the new generators to achieve similar generation quality to state-of-the-art convolutional generators. We also investigate several interesting properties unique to the new architecture.