Saved in:
Bibliographic Details
Main Authors: Lu, Mingjian, Tripathi, Pawan K., Shteyn, Mark, Ganguly, Debargha, French, Roger H., Chaudhary, Vipin, Wu, Yinghui
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.04154
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912875050696704
author Lu, Mingjian
Tripathi, Pawan K.
Shteyn, Mark
Ganguly, Debargha
French, Roger H.
Chaudhary, Vipin
Wu, Yinghui
author_facet Lu, Mingjian
Tripathi, Pawan K.
Shteyn, Mark
Ganguly, Debargha
French, Roger H.
Chaudhary, Vipin
Wu, Yinghui
contents Segmentation architectures are typically benchmarked on single imaging modalities, obscuring deployment-relevant performance variations: an architecture optimal for one modality may underperform on another. We present a cross-modal evaluation framework for materials image segmentation spanning SEM, AFM, XCT, and optical microscopy. Our evaluation of six encoder-decoder combinations across seven datasets reveals that optimal architectures vary systematically by context: UNet excels for high-contrast 2D imaging while DeepLabv3+ is preferred for the hardest cases. The framework also provides deployment feedback via out-of-distribution detection and counterfactual explanations that reveal which microstructural features drive predictions. Together, the architecture guidance, reliability signals, and interpretability tools address a practical gap in materials characterization, where researchers lack tools to select architectures for their specific imaging setup or assess when models can be trusted on new samples.
format Preprint
id arxiv_https___arxiv_org_abs_2602_04154
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Context Determines Optimal Architecture in Materials Segmentation
Lu, Mingjian
Tripathi, Pawan K.
Shteyn, Mark
Ganguly, Debargha
French, Roger H.
Chaudhary, Vipin
Wu, Yinghui
Computer Vision and Pattern Recognition
Segmentation architectures are typically benchmarked on single imaging modalities, obscuring deployment-relevant performance variations: an architecture optimal for one modality may underperform on another. We present a cross-modal evaluation framework for materials image segmentation spanning SEM, AFM, XCT, and optical microscopy. Our evaluation of six encoder-decoder combinations across seven datasets reveals that optimal architectures vary systematically by context: UNet excels for high-contrast 2D imaging while DeepLabv3+ is preferred for the hardest cases. The framework also provides deployment feedback via out-of-distribution detection and counterfactual explanations that reveal which microstructural features drive predictions. Together, the architecture guidance, reliability signals, and interpretability tools address a practical gap in materials characterization, where researchers lack tools to select architectures for their specific imaging setup or assess when models can be trusted on new samples.
title Context Determines Optimal Architecture in Materials Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2602.04154