Design-based CD and EPE metrology from low-dose SEM images: learned contours with deformable convolutions, bias-aware registration and a CUDA edge kernel. Exact-truth simulation plus two real SEM datasets, 17 pre-registered experiments.
cuda pytorch metrology u-net deformable-convolutional-networks defect-detection scanning-electron-microscopy computational-lithography cd-sem edge-placement-error semiconductor-metrology
-
Updated
Nov 16, 2026 - Python