NP-DRAW: A Non-Parametric Structured Latent Variable Model for Image Generation

TL;DR

NP-DRAW introduces a non-parametric, part-by-part image generation model using a latent canvas and Transformer dependency modeling, achieving superior performance and interpretability over previous structured models.

Contribution

The paper presents a novel non-parametric prior, Transformer-based dependency modeling, and a heuristic parsing algorithm for improved structured image generation.

Findings

Outperforms previous models like DRAW and AIR on multiple datasets.

Achieves competitive results with other generative models.

Enhances low-data learning and latent space interpretability.

Abstract

In this paper, we present a non-parametric structured latent variable model for image generation, called NP-DRAW, which sequentially draws on a latent canvas in a part-by-part fashion and then decodes the image from the canvas. Our key contributions are as follows. 1) We propose a non-parametric prior distribution over the appearance of image parts so that the latent variable ``what-to-draw'' per step becomes a categorical random variable. This improves the expressiveness and greatly eases the learning compared to Gaussians used in the literature. 2) We model the sequential dependency structure of parts via a Transformer, which is more powerful and easier to train compared to RNNs used in the literature. 3) We propose an effective heuristic parsing algorithm to pre-train the prior. Experiments on MNIST, Omniglot, CIFAR-10, and CelebA show that our method significantly outperforms previous structured image models like DRAW and AIR and is competitive to other generic generative models. Moreover, we show that our model's inherent compositionality and interpretability bring significant benefits in the low-data learning regime and latent space editing. Code is available at https://github.com/ZENGXH/NPDRAW.