Replace speculative tags with archive descriptors
Remove redundant or misleading tags (video, time-series, longitudinal-study, computer-vision, self-supervised-learning). Add literal descriptors for the published object: art, time-lapse, analog-drawing, drawing-process, longitudinal-archive, single-creator, provenance, and archival. Retain feature-extraction as the separate supported task facet.
Final scope (supersedes the opening description): keep video and computer-vision for routing; keep longitudinal-study and drawing-process as the distinctive intersection; remove only time-series and self-supervised-learning; add art, time-lapse, analog-media, skill-acquisition, single-creator, and provenance; retain archival. The README also clarifies that the July 15, 2025 acquisition-protocol boundary supports era-aware analysis but is not itself a skill-change label.
FINAL REVIEW SCOPE (supersedes earlier iteration notes): nine free tags in two layers. Accuracy: drawing-process, longitudinal-study, traditional-art, single-subject, video. Traffic: computer-vision, video-understanding, creative-process, human-motion. Removed: time-series and self-supervised-learning. Structured fields: task_categories video-classification + feature-extraction; annotations_creators expert-generated for creator-curated session/provenance metadata; language en for filenames, metadata, and documentation. The lead now foregrounds 467 recordings, one artist, physical media, nearly two years, and standardized acquisition since July 15, 2025. No license change.
Final verification passed: exact nine-tag taxonomy, structured facets, stronger lead, unchanged CC BY-NC 4.0 license, byte-identical upload, conflict-free PR, and README.md is the only changed path.