SegMet: Inverse Full LH-Surface Design using Classification-based Segmentation Algorithm
A CNN-based inverse design for full-metalens prediction, using segmentation and multi-class classification, achieves 2 × more accuracy than baseline Deep Learning methodologies. This reduces resource requirements, enabling precise beam shaping for customized focusing like live-cell imaging.
Authors
- Arnav Saha
Publication Details
- Published
- 2025-01-01
- DOI
- https://doi.org/10.1364/cleo_at.2025.jps100_40
- Primary Topic
- Advanced Numerical Analysis Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00