Context Determines Optimal Architecture in Materials Segmentation
This addresses a practical gap for materials researchers who lack tools to choose architectures for specific imaging setups or assess model trustworthiness on new samples, though it is incremental as it builds on existing methods.
The study tackled the problem of selecting optimal segmentation architectures for materials imaging by evaluating six encoder-decoder combinations across seven datasets, revealing that UNet excels for high-contrast 2D imaging while DeepLabv3+ is preferred for harder cases.
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.