Instance segmentation and microstructure characterization based on powder microscopic image data

doi:10.1016/j.matchar.2026.116051

Instance segmentation and microstructure characterization based on powder microscopic image data

2026-02-11

0

PRE

AI

DOI：10.1016/j.matchar.2026.116051

Original

Original request for help

Save

Abstract

En 中文

• YOLOv9-seg excels in IN718 powder segmentation, outperforming others. • Six parameters (area, sphericity etc.) reveal powder size-shape correlations. • Model maintains F1 > 0.94 on untested alloys (IN625/AgCuNi) without retraining. • Larger particles show higher sphericity and more regular morphology. • Sphericity identified as optimal shape descriptor with minimal outlier.

Keywords:

YOLOv9-seg

powder segmentation

sphericity

microstructure characterization

shape descriptors

Instance segmentation and microstructure characterization based on powder microscopic image data

Abstract

Journal

Organization