A multimodal fiber reinforced polymer composite micrograph dataset for image-based analysis benchmark

This record supplements the article Benchmarking and sensitivity analysis of segmentation methods for image-based fiber detection in composites, doi:10.1016/j.compositesa.2026.110150. Refer to the supplemented paper for the detailed methodology. Content Three micrographs of the same polished cross-section of a unidirectional carbon-fiber/LM-PAEK laminate ([0]16, T700G fiber, TC1225 prepreg, Toray Advanced Composites), acquired on a Keyence VK-X1000 confocal microscope at three magnification/objective conditions: 50X (working distance 1.1 mm) — pixel density 3615 px/mm (0.277 µm/px) 20X Plan (working distance 3.1 mm) — pixel density 1446 px/mm (0.692 µm/px) 20X ELWD (working distance 11.0 mm) — pixel density 1446 px/mm (0.692 µm/px) Eleven research groups from eight institutions (anonymized as participants 1-11) independently processed these micrographs to detect fiber centroids and radii using their own algorithms; results are consolidated in benchmark_results/detected_fibers.csv. A single 190 x 100 µm subframe (selected for the highest cross-participant disagreement in fiber volume fraction) was manually annotated by three independent domain experts by ellipse-fitting in Fiji/ImageJ, providing an expert reference (not a validated ground truth) in annotated_data. Units Fiber centers and radii are reported in image pixels. To convert to millimeters: - 50X images: value_mm = value_px / 3615 - 20X images: value_mm = value_px / 1446 Full column definitions and units are given in the accompanying *_data_dictionary.json files. Files data Full stitched 32-bit RGB TIFF micrographs (uncorrected) benchmark_results Consolidated participant fiber detections (pixels) annotated_data Expert annotations (pixels) and crops of the reference subframe README.md Full technical documentation Contributors The authors acknowledge the contributors who submitted their results to this benchmark. Refer to the detailed list of authorship and contributions of the supplemented paper. G. Broggi and O. Yuksel contributed equally to this work. Citation Please cite both this dataset (via its DOI) and the associated article if appropriate:Benchmarking and sensitivity analysis of segmentation methods for image-based fiber detection in composites, doi:10.1016/j.compositesa.2026.110150.