Precision at Scale: Domain-Specific Datasets On-Demand

TL;DR

This paper introduces Precision at Scale (PaS), a method for automatically generating domain-specific datasets on-demand, which outperform traditional datasets like ImageNet in training visual models across various domains.

Contribution

The paper presents a novel modular pipeline for creating large, diverse, domain-specific datasets automatically, reducing reliance on human-labeled data and improving model pretraining performance.

Findings

PaS-generated datasets outperform traditional domain datasets in diversity and scale.

Models pretrained on PaS datasets outperform ImageNet-1k in multiple domains.

PaS datasets lead to better domain-specific model performance with smaller data sizes.

Abstract

In the realm of self-supervised learning (SSL), conventional wisdom has gravitated towards the utility of massive, general domain datasets for pretraining robust backbones. In this paper, we challenge this idea by exploring if it is possible to bridge the scale between general-domain datasets and (traditionally smaller) domain-specific datasets to reduce the current performance gap. More specifically, we propose Precision at Scale (PaS), a novel method for the autonomous creation of domain-specific datasets on-demand. The modularity of the PaS pipeline enables leveraging state-of-the-art foundational and generative models to create a collection of images of any given size belonging to any given domain with minimal human intervention. Extensive analysis in two complex domains, proves the superiority of PaS datasets over existing traditional domain-specific datasets in terms of diversity, scale, and effectiveness in training visual transformers and convolutional neural networks. Most notably, we prove that automatically generated domain-specific datasets lead to better pretraining than large-scale supervised datasets such as ImageNet-1k and ImageNet-21k. Concretely, models trained on domain-specific datasets constructed by PaS pipeline, beat ImageNet-1k pretrained backbones by at least 12% in all the considered domains and classification tasks and lead to better food domain performance than supervised ImageNet-21k pretrain while being 12 times smaller. Code repository: https://github.com/jesusmolrdv/Precision-at-Scale/