Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SegSEM: Enabling and Enhancing SAM2 for SEM Contour Extraction

Da Chen, Guangyu Hu, Kaihong Xu, Kaichao Liang, Songjiang Li, Wei Yang, XiangYu Wen, Mingxuan Yuan

TL;DR

SegSEM is proposed, a framework built on two principles: a data-efficient fine-tuning strategy that adapts by selectively training only the model's encoders, and a robust hybrid architecture integrating a traditional algorithm as a confidence-aware fallback.

Abstract

Extracting high-fidelity 2D contours from Scanning Electron Microscope (SEM) images is critical for calibrating Optical Proximity Correction (OPC) models. While foundation models like Segment Anything 2 (SAM2) are promising, adapting them to specialized domains with scarce annotated data is a major challenge. This paper presents a case study on adapting SAM2 for SEM contour extraction in a few-shot setting. We propose SegSEM, a framework built on two principles: a data-efficient fine-tuning strategy that adapts by selectively training only the model's encoders, and a robust hybrid architecture integrating a traditional algorithm as a confidence-aware fallback. Using a small dataset of 60 production images, our experiments demonstrate this methodology's viability. The primary contribution is a methodology for leveraging foundation models in data-constrained industrial applications.

SegSEM: Enabling and Enhancing SAM2 for SEM Contour Extraction

TL;DR

SegSEM is proposed, a framework built on two principles: a data-efficient fine-tuning strategy that adapts by selectively training only the model's encoders, and a robust hybrid architecture integrating a traditional algorithm as a confidence-aware fallback.

Abstract

Extracting high-fidelity 2D contours from Scanning Electron Microscope (SEM) images is critical for calibrating Optical Proximity Correction (OPC) models. While foundation models like Segment Anything 2 (SAM2) are promising, adapting them to specialized domains with scarce annotated data is a major challenge. This paper presents a case study on adapting SAM2 for SEM contour extraction in a few-shot setting. We propose SegSEM, a framework built on two principles: a data-efficient fine-tuning strategy that adapts by selectively training only the model's encoders, and a robust hybrid architecture integrating a traditional algorithm as a confidence-aware fallback. Using a small dataset of 60 production images, our experiments demonstrate this methodology's viability. The primary contribution is a methodology for leveraging foundation models in data-constrained industrial applications.
Paper Structure (13 sections, 1 equation, 6 figures, 1 table)

This paper contains 13 sections, 1 equation, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. The SegSEM Framework
  5. Core Component: Data-Efficient Adaptation of SAM2
  6. Selective Fine-Tuning Strategy
  7. Custom Loss Function for High-Fidelity Masks
  8. Ensuring Reliability: The Hybrid Architecture with Fallback Module
  9. Automated Mask Quality Assessment and Fallback
  10. Final Contour Generation
  11. Experiments and Results
  12. Experimental Setup and Results
  13. Conclusion

Figures (6)

  • Figure 1: (a) Optical Proximity Correction (OPC) standard procedure; (b) using SEM image to calibrate OPC model
  • Figure 2: The overview of traditional SEM contour extraction method
  • Figure 3: The overview of SEM contour extraction method based on SAM2 network and enhanced mask extractor (Left: Fine-tuned SAM2 model with unfrozen layers. Right: Automatic algorithm when SAM2 fails to extract an effective mask.)
  • Figure 4: Training process of the SAM2 model
  • Figure 5: Example of fallback activation: (a) Failure of SAM2 alone; (b) Successful extraction by the fallback module.
  • ...and 1 more figures