HeteroViT shows a single-layer ViT can handle both supervised hit classification and self-supervised rare-event detection across detector modalities while mapping cleanly to heterogeneous detector hardware for edge keep/discard decisions.
Rapid detection of rare events from in situ X-ray diffraction data using machine learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
physics.ins-det 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
HeteroViT: A Versatile Single-Layer Vision Transformer Concept, Co-Designed for Distributed Real-Time Data Reduction on Scientific Detectors
HeteroViT shows a single-layer ViT can handle both supervised hit classification and self-supervised rare-event detection across detector modalities while mapping cleanly to heterogeneous detector hardware for edge keep/discard decisions.