Learning to See by Looking at Noise

TL;DR

This paper explores training visual models solely on noise-generated images, demonstrating that certain structural properties and diversity in noise can enable effective learning without real datasets.

Contribution

It introduces a novel approach of using purely noise-based images for training visual representations, bypassing the need for real image datasets.

Findings

Structural properties in noise improve learning

Diversity in noise enhances representation quality

Good performance achieved with non-realistic noise processes

Abstract

Current vision systems are trained on huge datasets, and these datasets come with costs: curation is expensive, they inherit human biases, and there are concerns over privacy and usage rights. To counter these costs, interest has surged in learning from cheaper data sources, such as unlabeled images. In this paper we go a step further and ask if we can do away with real image datasets entirely, instead learning from noise processes. We investigate a suite of image generation models that produce images from simple random processes. These are then used as training data for a visual representation learner with a contrastive loss. We study two types of noise processes, statistical image models and deep generative models under different random initializations. Our findings show that it is important for the noise to capture certain structural properties of real data but that good performance can be achieved even with processes that are far from realistic. We also find that diversity is a key property to learn good representations. Datasets, models, and code are available at https://mbaradad.github.io/learning_with_noise.