Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Segment Anything Model for Grain Characterization in Hard Drive Design

Kai Nichols, Matthew Hauwiller, Nicholas Propes, Shaowei Wu, Stephanie Hernandez, Mike Kautzky

TL;DR

Hard disk drive design requires nanoscale grain segmentation to understand materials for HAMR. The paper evaluates Meta's Segment Anything Model (SAM) and its Automatic Mask Generator (AMG) for zero-shot grain segmentation on SEM images, and investigates training-free and training-based avenues to improve performance with minimal labeled data. The study uses five SEM images of a gold film (783 hand-traced grains) to compare SAM-derived masks and hand-labeled grains, analyzing grain properties like area, elongatedness, and perimeter. Results show promising out-of-the-box property extraction but reveal systematic biases and missed grains, with iterative prompting and domain-aware post-processing offering meaningful, training-free gains, while future work should pursue data augmentation, style transfer, and weakly supervised pre-training to extend applicability.

Abstract

Development of new materials in hard drive designs requires characterization of nanoscale materials through grain segmentation. The high-throughput quickly changing research environment makes zero-shot generalization an incredibly desirable feature. For this reason, we explore the application of Meta's Segment Anything Model (SAM) to this problem. We first analyze the out-of-the-box use of SAM. Then we discuss opportunities and strategies for improvement under the assumption of minimal labeled data availability. Out-of-the-box SAM shows promising accuracy at property distribution extraction. We are able to identify four potential areas for improvement and show preliminary gains in two of the four areas.

Segment Anything Model for Grain Characterization in Hard Drive Design

TL;DR

Hard disk drive design requires nanoscale grain segmentation to understand materials for HAMR. The paper evaluates Meta's Segment Anything Model (SAM) and its Automatic Mask Generator (AMG) for zero-shot grain segmentation on SEM images, and investigates training-free and training-based avenues to improve performance with minimal labeled data. The study uses five SEM images of a gold film (783 hand-traced grains) to compare SAM-derived masks and hand-labeled grains, analyzing grain properties like area, elongatedness, and perimeter. Results show promising out-of-the-box property extraction but reveal systematic biases and missed grains, with iterative prompting and domain-aware post-processing offering meaningful, training-free gains, while future work should pursue data augmentation, style transfer, and weakly supervised pre-training to extend applicability.

Abstract

Development of new materials in hard drive designs requires characterization of nanoscale materials through grain segmentation. The high-throughput quickly changing research environment makes zero-shot generalization an incredibly desirable feature. For this reason, we explore the application of Meta's Segment Anything Model (SAM) to this problem. We first analyze the out-of-the-box use of SAM. Then we discuss opportunities and strategies for improvement under the assumption of minimal labeled data availability. Out-of-the-box SAM shows promising accuracy at property distribution extraction. We are able to identify four potential areas for improvement and show preliminary gains in two of the four areas.
Paper Structure (11 sections, 2 figures, 3 tables)

This paper contains 11 sections, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Evaluating SAM for Grain Characterization
  3. Data
  4. Segmentation Method
  5. Property Extraction
  6. SAM's Limitations
  7. Training-Free Improvements
  8. Prompt-Selection
  9. Post-Processing
  10. Training-Based Improvements
  11. Conclusion

Figures (2)

  • Figure 1: Segmentation with Automatic Mask Generator. (a) SEM images. (b) Hand drawn tracings. (c) SAM AMG + mask-level NMS segmentation. Masks are white with red outline. Where no masks were found is black. (d-f) Density Histograms of microstructure properties for hand-labeled grains (orange) and predicted grains (blue).
  • Figure 2: Comparison of ability of SAM to capture hand-labeled grains under increasingly generous conditions.