A 22,266-Image Grazing-Incidence Wide-Angle X-ray Scattering Benchmark Dataset for Organic Photovoltaic and Perovskite Materials Characterization

Curated dataset The main curated dataset is provided as: GIWAXS_PVK_OPV_Dataset.zip This archive contains the processed dataset of 22,266 GIWAXS patterns from organic photovoltaic and perovskite thin films, measured at the SAXS/WAXS beamline of the Australian Synchrotron. It also includes utilities for data loading and filename-based label parsing, a ResNet-50 classification baseline, a retrieval-augmented generation (RAG-LLM) example for retrieval-grounded GIWAXS comparison and interpretation, and a transfer-learning example demonstrating cross-modality learning with X-ray diffraction data. Raw experimental data The following archives contain the original raw experimental data from which the curated dataset was generated: GIWAXS_2024_experiments.7z GIWAXS_Nov_2025_experiments.7z GIWAXS_2026_experiments.7z GIWAXS_2026_experiments_images_Part01.7z GIWAXS_2026_experiments_images_Part02.7z GIWAXS_2026_experiments_images_Part03.7z GIWAXS_2026_experiments_images_Part04.7z GIWAXS_2026_experiments_images_Part05.7z GIWAXS_2026_experiments_images_Part06.7z GIWAXS_2026_experiments_images_Part07.7z These raw-data archives are provided separately from the curated dataset to preserve access to the underlying experimental measurements. For the 2026 raw experimental data, the images directory was too large to distribute as a single archive. It was therefore removed from GIWAXS_2026_experiments.7z and divided across the independently extractable GIWAXS_2026_experiments_images_PartXX.7z archives. Reconstructing the 2026 raw-data directory First extract: GIWAXS_2026_experiments.7z Then extract all: GIWAXS_2026_experiments_images_PartXX.7z Combine the contents of the extracted images folders into: GIWAXS_march_2026/exp1/images/ All image files from Part01, Part02, etc. belong to the same original images directory. The PartXX designation is used only for file distribution and does not indicate separate experiments, classes, or subsets of the dataset.