README.md

metadata

pretty_name: ALD-E-ImageMiner
language:
  - en
annotations_creators:
  - expert-generated
  - machine-generated
language_creators:
  - found
multilinguality:
  - monolingual
source_datasets:
  - original
task_categories:
  - image-classification
  - image-to-text
  - visual-question-answering
size_categories:
  - 1K<n<10K
tags:
  - scientific-images
  - materials-science
  - scientific-figures
  - atomic-layer-deposition
  - atomic-layer-etching
  - experimental-usecase
  - simulation-usecase
  - multimodal
  - image
  - text
  - imagefolder
  - datasets
  - arxiv:2607.26848
license: other
license_name: mixed-rights-non-commercial
license_link: LICENSE
configs:
  - config_name: molecular-structure-diagram
    data_dir: subsets/molecular-structure-diagram
    default: true
  - config_name: 3d-scatter-plot
    data_dir: subsets/3d-scatter-plot
  - config_name: apparatus-diagram
    data_dir: subsets/apparatus-diagram
  - config_name: area-chart
    data_dir: subsets/area-chart
  - config_name: band-diagram
    data_dir: subsets/band-diagram
  - config_name: bar-chart
    data_dir: subsets/bar-chart
  - config_name: box-plot
    data_dir: subsets/box-plot
  - config_name: chromaticity-diagram
    data_dir: subsets/chromaticity-diagram
  - config_name: conceptual-diagram
    data_dir: subsets/conceptual-diagram
  - config_name: contour-heatmap
    data_dir: subsets/contour-heatmap
  - config_name: device-structure-diagram
    data_dir: subsets/device-structure-diagram
  - config_name: formula
    data_dir: subsets/formula
  - config_name: grouped-bar-chart
    data_dir: subsets/grouped-bar-chart
  - config_name: heatmap
    data_dir: subsets/heatmap
  - config_name: image-panel
    data_dir: subsets/image-panel
  - config_name: line-chart
    data_dir: subsets/line-chart
  - config_name: multi-axis-chart
    data_dir: subsets/multi-axis-chart
  - config_name: multi-spectra-chart
    data_dir: subsets/multi-spectra-chart
  - config_name: multiple-line-chart
    data_dir: subsets/multiple-line-chart
  - config_name: multiple-scatter-plot
    data_dir: subsets/multiple-scatter-plot
  - config_name: network-diagram
    data_dir: subsets/network-diagram
  - config_name: periodic-table-map
    data_dir: subsets/periodic-table-map
  - config_name: phase-diagram
    data_dir: subsets/phase-diagram
  - config_name: pie-chart
    data_dir: subsets/pie-chart
  - config_name: polar-chart-rose-chart
    data_dir: subsets/polar-chart-rose-chart
  - config_name: process-flow-diagram
    data_dir: subsets/process-flow-diagram
  - config_name: process-timing-diagram
    data_dir: subsets/process-timing-diagram
  - config_name: reaction-energy-profile-diagram
    data_dir: subsets/reaction-energy-profile-diagram
  - config_name: reaction-scheme
    data_dir: subsets/reaction-scheme
  - config_name: scatter-plot
    data_dir: subsets/scatter-plot
  - config_name: spectra-chart
    data_dir: subsets/spectra-chart
  - config_name: stacked-bar-chart
    data_dir: subsets/stacked-bar-chart
  - config_name: stacked-spectra-chart
    data_dir: subsets/stacked-spectra-chart
  - config_name: table
    data_dir: subsets/table
  - config_name: timeline-chart
    data_dir: subsets/timeline-chart
  - config_name: unknown
    data_dir: subsets/unknown
  - config_name: workflow-diagram
    data_dir: subsets/workflow-diagram

ALD-E-ImageMiner

ALD-E-ImageMiner is a benchmark package for scientific figure understanding in atomic layer deposition and atomic layer etching literature. This Hugging Face export is organized for ImageFolder loading while preserving the repository's source split membership and panel-level annotations.

🗂️ Source Data Summary

This package was generated from the sciknoworg/ALD-E-ImageMiner GitHub repository, using icdar2026-competition-data, and contains 1951 full-figure image records.

Hugging Face split	Source split	Records
train	train	1170
validation	dev	201
test	test/gold_standard_test_set	580

The source trial, test/blind_test_set, and test/submission_guidelines directories are not included.

🧭 Subsets And Splits

Each Hugging Face subset/config is derived from the source classification annotation. When a full figure contains multiple panel classifications, this package uses the first classification label in source JSON order as the subset assignment and keeps the complete panel classification list in the classification column.

Available subsets (37):

3d-scatter-plot, apparatus-diagram, area-chart, band-diagram, bar-chart, box-plot, chromaticity-diagram, conceptual-diagram, contour-heatmap, device-structure-diagram, formula, grouped-bar-chart, heatmap, image-panel, line-chart, molecular-structure-diagram, multi-axis-chart, multi-spectra-chart, multiple-line-chart, multiple-scatter-plot, network-diagram, periodic-table-map, phase-diagram, pie-chart, polar-chart-rose-chart, process-flow-diagram, process-timing-diagram, reaction-energy-profile-diagram, reaction-scheme, scatter-plot, spectra-chart, stacked-bar-chart, stacked-spectra-chart, table, timeline-chart, unknown, workflow-diagram

The source dev split is exposed as the Hugging Face validation split.

🧾 Schema

Column	Description
file_name	Relative path to the copied image file. This is first so ImageFolder renders the image column.
caption	Caption extracted from the source content.json when available.
id	Stable package identifier using the Hugging Face split and source sample_id.
sample_id	Original source sample_id.
subset	Lowercase kebab-case subset chosen from the first source classification label.
split	Hugging Face split: train, validation, or test.
classification	All source panel classifications as a list of panel_id/label structs.
summarization, data_extraction, vqa	Source task annotations preserved as JSON strings containing stable panel lists. Empty lists indicate that the source record did not provide that task annotation.
bbox	Source bounding boxes converted to a stable panel list.
source	Relative path to the source image in this repository.
provenance	Relative annotation/content/PDF paths and source navigation fields.
width, height, image_format, image_sha256	Image properties computed from the copied source image.
metadata_license, image_license, image_reuse_status	Conservative reuse and licensing metadata.
schema_version	Metadata schema version, currently 1.0.0.

🚀 Usage

from datasets import load_dataset

dataset = load_dataset("SciKnowOrg/ALD-E-ImageMiner", "molecular-structure-diagram")
train = dataset["train"]

For a local checkout of this generated package:

from datasets import load_dataset

dataset = load_dataset("imagefolder", data_dir="hf/dataset/subsets/molecular-structure-diagram")

⚖️ License And Reuse

This package is marked as mixed-rights-non-commercial because the figure images come from many scientific articles with article-specific reuse terms, and this release is intended for non-commercial research, benchmarking, and evaluation use.

Dataset annotations and generated metadata are released under CC BY 4.0 unless a more specific file-level notice says otherwise. For images, the reuse granularity is record/image-level rather than corpus-level: each extracted figure image and source-derived paper field follows the rights and reuse terms of its corresponding source article. This package does not grant commercial reuse rights for the images or source-paper content.

The per-record license fields use conservative values: metadata_license is CC BY 4.0, image_license is source_publisher_rights_reserved, and image_reuse_status is non_commercial_research_use_only. Check the LICENSE, source, and provenance fields before reusing or redistributing any image.

📖 Citation

The vision working paper for this project is pre-released on Zenodo. Please cite this paper if you find the project useful:

@misc{d_souza_2025_17130928,
  author       = {D'Souza, Jennifer},
  title        = {A Pathway to General-Purpose Scientific AI:
                   Multimodal Comprehension of Scientific Images},
  month        = sep,
  year         = 2025,
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.17130928},
  url          = {https://doi.org/10.5281/zenodo.17130928},
}

This benchmark dataset was used as the ICDAR 2026 Competition on Information Extraction from Atomic Layer Deposition/Etching (ALD/E) Scientific Figures, organized as part of the ICDAR 2026 competitions. Please cite the competition report when referring to the ALD/E-ImageMiner benchmark, dataset, competition, or associated information-extraction tasks:

@article{ahmed2026icdar,
  title   = {ICDAR 2026 Competition on Information Extraction from
             Atomic Layer Deposition/Etching (ALD/E) Scientific Figures},
  author  = {Ahmed, Fahad and Auer, S{\"o}ren and D'Souza, Jennifer},
  journal = {arXiv preprint arXiv:2607.26848},
  year    = {2026},
  url     = {https://arxiv.org/abs/2607.26848}
}

⭐ Acknowledgements

The development of the expert-annotated ALD/E-ImageMiner benchmark resource was funded by the NFDI4DataScience initiative, funded by the German Research Foundation (DFG, Grant ID: 460234259) under the Speedboat Annotation Project funding scheme.

This research endeavor is conducted in the context of the AI-Aware Pathways to Sustainable Semiconductor Process and Manufacturing Technologies (AWASES) initiative (Mackus et al., 2024), funded by Merck and Intel, with collaboration between Eindhoven University, Leibniz University Hannover's L3S Research Centre, and University of Warwick. AWASES hosts three fully funded PhD positions and supports advances in generative AI, multimodal models, and FAIR scientific knowledge graph construction.