Better visual data
starts with context.
Scale alone does not produce visual understanding. INsVI explores smaller, intentionally curated datasets that preserve the histories, processes, relationships, and judgments surrounding images.
Images are not interchangeable files.
Provenance
Document source, creator, date, rights, transformations, and collection history.
Visual analysis
Record composition, material, process, spatial structure, and salient relationships.
Cultural context
Preserve interpretation, uncertainty, contested meaning, and limits of annotation.
Quality is a research decision.
A visually meaningful dataset requires choices about inclusion, comparison, sequencing, metadata, expert review, and intended use.
Traceability
Make the origin and processing history of each item inspectable.
Plural interpretation
Avoid presenting a single caption or classification as the image’s complete meaning.
Responsible access
Match licensing, consent, attribution, and access conditions to the material.
A framework in progress.
INsVI is developing principles and prototype workflows for visual-data curation. Dataset releases will appear here only when scope, rights, documentation, and methods are ready for public use.