Transfer learning with DinoV2 and pseudo-labeling achieves over 90% defect classification accuracy on semiconductor SEM images using fewer than 15 labeled images per class.
DeepSEM -Net: Enhancing SEM defect analysis in semiconductor manufacturing with a dual -branch CNN -Transformer architecture
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Semiconductor SEM Image Defect Classification Using Supervised and Semi-Supervised Learning with Vision Transformers
Transfer learning with DinoV2 and pseudo-labeling achieves over 90% defect classification accuracy on semiconductor SEM images using fewer than 15 labeled images per class.