A synthetic physical inconsistency detector using DINOv2 residual probes, multi-view validation, temporal sync drift checks, semantic mismatch detection, and counterfactual intervention generation.
computer-vision pytorch active-learning anomaly-detection rerun synthetic-data multiview-learning synthetic-dataset-generation mlops semantic-reasoning dinov2 temporal-validation self-healing-data
-
Updated
Jun 9, 2026 - Python