PixelNN: Example-based Image Synthesis

TL;DR

PixelNN introduces a hybrid approach combining CNNs and nearest neighbor methods to generate diverse, high-quality, photorealistic images from incomplete signals, addressing limitations of existing deep generative models.

Contribution

The paper presents a simple pipeline that integrates CNNs with pixel-wise nearest neighbor search to improve diversity and interpretability in image synthesis from incomplete inputs.

Findings

Achieves diverse high-frequency image outputs from various incomplete signals.

Outperforms traditional generative models in diversity and controllability.

Effective across multiple domains like faces, animals, and objects.

Abstract

We present a simple nearest-neighbor (NN) approach that synthesizes high-frequency photorealistic images from an "incomplete" signal such as a low-resolution image, a surface normal map, or edges. Current state-of-the-art deep generative models designed for such conditional image synthesis lack two important things: (1) they are unable to generate a large set of diverse outputs, due to the mode collapse problem. (2) they are not interpretable, making it difficult to control the synthesized output. We demonstrate that NN approaches potentially address such limitations, but suffer in accuracy on small datasets. We design a simple pipeline that combines the best of both worlds: the first stage uses a convolutional neural network (CNN) to maps the input to a (overly-smoothed) image, and the second stage uses a pixel-wise nearest neighbor method to map the smoothed output to multiple high-quality, high-frequency outputs in a controllable manner. We demonstrate our approach for various input modalities, and for various domains ranging from human faces to cats-and-dogs to shoes and handbags.