Dataset Factory: A Toolchain For Generative Computer Vision Datasets
This addresses data wrangling and versioning issues for machine learning practitioners working with petabyte-scale datasets, though it appears incremental as it builds on existing data management concepts.
The paper tackles the challenges of managing large-scale generative computer vision datasets by proposing a dataset factory approach that separates sample storage from metadata processing, enabling scalable data-centric operations for ML teams and researchers.
Generative AI workflows heavily rely on data-centric tasks - such as filtering samples by annotation fields, vector distances, or scores produced by custom classifiers. At the same time, computer vision datasets are quickly approaching petabyte volumes, rendering data wrangling difficult. In addition, the iterative nature of data preparation necessitates robust dataset sharing and versioning mechanisms, both of which are hard to implement ad-hoc. To solve these challenges, we propose a "dataset factory" approach that separates the storage and processing of samples from metadata and enables data-centric operations at scale for machine learning teams and individual researchers.