Find research datasets worth reusing
Search datasets from major research repositories and use ShareScore to quickly assess how well each record supports discovery, access, and reuse.
155
datasets available to search
ShareScore release 0.9.0
Dataset results
155 results for “2D images”
[MedMNIST+] 18x Standardized Datasets for 2D and 3D Biomedical Image Classification with Multiple Size Options: 28 (MNIST-Like), 64, 128, and 224
<h2><strong>Code</strong> [<a href="https://github.com/MedMNIST/MedMNIST" target="_blank" rel="noopener">GitHub</a>] | <strong>Publication</strong> [<a href="https://doi.org/10.1038/s41597-022-01721-8" target="_blank" rel="noopener">Nature Scientific Data'23</a> / <a href="https://doi.org/10.1109/ISBI48211.2021.9434062" target="_blank" rel="noopener">ISBI'21</a>] | <strong>Preprint</strong> [<a href="https://arxiv.org/abs/2110.14795" target="_blank" rel="noopener">arXiv</a>]</h2> <p> </p> <p><strong>Abstract</strong></p> <p>We introduce MedMNIST, a large-scale MNIST-like collection of standardized biomedical images, including 12 datasets for 2D and 6 datasets for 3D. All images are pre-processed into 28x28 (2D) or 28x28x28 (3D) with the corresponding classification labels, so that no background knowledge is required for users. Covering primary data modalities in biomedical images, MedMNIST is designed to perform classification on lightweight 2D and 3D images with various data scales (from 100 to 100,000) and diverse tasks (binary/multi-class, ordinal regression and multi-label). The resulting dataset, consisting of approximately 708K 2D images and 10K 3D images in total, could support numerous research and educational purposes in biomedical image analysis, computer vision and machine learning. We benchmark several baseline methods on MedMNIST, including 2D / 3D neural networks and open-source / commercial AutoML tools. The data and code are publicly available at <a href="https://medmnist.com/">https://medmnist.com/</a>.</p> <p><em><strong>Disclaimer</strong></em>: The only official distribution link for the MedMNIST dataset is <a href="https://doi.org/10.5281/zenodo.10519652">Zenodo</a>. We kindly request users to refer to this original dataset link for accurate and up-to-date data.</p> <p><strong><em>Update</em>:</strong> We are thrilled to release <a href="https://github.com/MedMNIST/MedMNIST/blob/main/on_medmnist_plus.md">MedMNIST+</a> with larger sizes: 64x64, 128x128, and 224x224 for 2D, and 64x64x64 for 3D. As a complement to the previous 28-size MedMNIST, the large-size version could serve as a standardized benchmark for medical foundation models. Install the latest API to try it out!</p> <p> </p> <p><strong>Python Usage</strong></p> <p>We recommend our official <a href="https://github.com/MedMNIST/MedMNIST">code</a> to download, parse and use the MedMNIST dataset:</p> <blockquote> <pre>% pip install medmnist<br>% python</pre> <div> <div>To use the standard 28-size (MNIST-like) version utilizing the downloaded files:</div> <br> <div>>>> from medmnist import PathMNIST</div> <div>>>> train_dataset = PathMNIST(split="train")</div> <br> <div>To enable automatic downloading by setting `download=True`:</div> <br> <div>>>> from medmnist import NoduleMNIST3D</div> <div>>>> val_dataset = NoduleMNIST3D(split="val", download=True)</div> <br> <div>Alternatively, you can access MedMNIST+ with larger image sizes by specifying the `size` parameter:</div> <br> <div>>>> from medmnist import ChestMNIST</div> <div>>>> test_dataset = ChestMNIST(split="test", download=True, size=224)</div> </div> </blockquote> <p> </p> <p><strong>Citation</strong></p> <p>If you find this project useful, please cite both v1 and v2 paper as:</p> <blockquote> <p>Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, Bingbing Ni. Yang, Jiancheng, et al. "MedMNIST v2-A large-scale lightweight benchmark for 2D and 3D biomedical image classification." Scientific Data, 2023.</p> <p>Jiancheng Yang, Rui Shi, Bingbing Ni. "MedMNIST Classification Decathlon: A Lightweight AutoML Benchmark for Medical Image Analysis". IEEE 18th International Symposium on Biomedical Imaging (ISBI), 2021.</p> </blockquote> <p>or using bibtex:</p> <blockquote> <pre>@article{medmnistv2, title={MedMNIST v2-A large-scale lightweight benchmark for 2D and 3D biomedical image classification}, author={Yang, Jiancheng and Shi, Rui and Wei, Donglai and Liu, Zequan and Zhao, Lin and Ke, Bilian and Pfister, Hanspeter and Ni, Bingbing}, journal={Scientific Data}, volume={10}, number={1}, pages={41}, year={2023}, publisher={Nature Publishing Group UK London} } @inproceedings{medmnistv1, title={MedMNIST Classification Decathlon: A Lightweight AutoML Benchmark for Medical Image Analysis}, author={Yang, Jiancheng and Shi, Rui and Ni, Bingbing}, booktitle={IEEE 18th International Symposium on Biomedical Imaging (ISBI)}, pages={191--195}, year={2021} }</pre> </blockquote> <p>Please also cite the corresponding paper(s) of source data if you use any subset of MedMNIST as per the description on the <a href="https://medmnist.github.io/">project website</a>.</p> <p> </p> <p><strong>License</strong></p> <p>The MedMNIST dataset is licensed under <em>Creative Commons Attribution 4.0 International</em> (<a href="https://creativecommons.org/licenses/by/4.0/">CC BY 4.0</a>), except DermaMNIST under <em>Creative Commons Attribution-NonCommercial 4.0 International</em> (<a href="https://creativecommons.org/licenses/by-nc/4.0/">CC BY-NC 4.0</a>).</p> <p>The code is under <a href="https://github.com/MedMNIST/MedMNIST/blob/main/LICENSE">Apache-2.0 License</a>.</p> <p> </p> <p><strong>Changelog</strong></p> <p><a href="https://doi.org/10.5281/zenodo.10519652">v3.0</a> (this repository): Released MedMNIST+ featuring larger sizes: 64x64, 128x128, and 224x224 for 2D, and 64x64x64 for 3D.</p> <p><a href="https://doi.org/10.5281/zenodo.10519195">v2.2</a>: Removed a small number of mistakenly included blank samples in OrganAMNIST, OrganCMNIST, OrganSMNIST, OrganMNIST3D, and VesselMNIST3D. </p> <p><a href="https://doi.org/10.5281/zenodo.6496656">v2.1</a>: Addressed an issue in the NoduleMNIST3D file (i.e., nodulemnist3d.npz). Further details can be found in this <a href="https://github.com/MedMNIST/MedMNIST/issues/22#issuecomment-1103438191">issue</a>.</p> <p><a href="https://doi.org/10.5281/zenodo.5208230">v2.0</a>: Launched the initial repository of MedMNIST v2, adding 6 datasets for 3D and 2 for 2D.</p> <p><a href="https://doi.org/10.5281/zenodo.4269852">v1.0</a>: Established the initial repository (in a separate repository) of MedMNIST v1, featuring 10 datasets for 2D.</p> <p> </p> <p><strong>Note</strong>: This dataset is <strong>NOT</strong> intended for clinical use.</p>
Digital Repository of Ireland Member Digitisation Workflows for 2D Image Files: Survey Questions and Dataset
<p>The Digital Repository of Ireland (DRI) issued a survey to its membership, <strong>DRI Member Digitisation Workflows for 2D Images</strong>, which ran from December 7, 2023–January 31, 2024. The survey was conducted to improve the DRI’s understanding of the technical processes and metadata workflows that our members use to digitise and share images in the Repository, in order to better tailor our support for this work and deliver the most complete information about digital images files available to our users. </p> <p>The survey informed the actions taken in WorldFAIR Project WP13 deliverable <a href="https://doi.org/10.5281/zenodo.10850009" target="_blank" rel="noopener">13.3 Implementing and Testing the Cultural Heritage Image Sharing Recommendations: DRI Case Study Report</a>. The data will inform ongoing work at DRI aimed at improving the transparency of technical information associated with digital assets accessed through the Repository.</p> <p>Read more about the Cultural Heritage Image Sharing Case Study DRI on our website: <a href="https://dri.ie/the-worldfair-project/">https://dri.ie/the-worldfair-project/</a>. </p> <p>Summary: DRI is Ireland's national repository for the arts, humanities, and social sciences data, and operates on a membership scheme. There were 20 respondents to the survey, giving us a response rate of about 35% of DRI's membership. Representation from professional fields of work across the cultural heritage sector was captured in the results (note that some institutions gave multiple responses): 17 Archives, 12 Libraries, 5 Museums and 11 Higher Education Institutions. </p>
In silico 2D photoacoustic imaging data
<p>Here you find the data that was used for the experiments in the paper <strong>Confidence estimation for machine learning-based quantitative photoacoustics</strong> by <em>Janek Gröhl</em>, <em>Thomas Kirchner</em>, <em>Tim Adler</em>, and <em>Lena Maier-Hei</em>n.</p>
3D IQ Test Task (3D-IQTT) - A Dataset for Quantitative Evaluation of 3D Reconstruction from 2D Images
<p>3D reconstruction is mostly evaluated qualitatively. With this dataset, we are introducing a new difficult quantitative task, the 3D IQ test task (3D-IQTT).</p> <p>It is designed to be similar to mental rotation questions found in some IQ tests. Each element in the dataset consists of 4 images: reference object and answers 1-3. One of the answers is the reference object but randomly rotated. For every question, dataset users have to use their model to pick the rotated model out of the 3 possible answers.</p> <p>The dataset encourages semi-supervised or unsupervised 3D reconstruction because it contains a large corpus of unlabeled data and only a small set of labeled data where the correct answer is known.</p> <p>All the images are of blocky 3D shapes floating in space in front of a black background.</p> <p>Demo scripts for loading/processing the dataset can be found at <a href="https://github.com/fgolemo/3D-IQTT">https://github.com/fgolemo/3D-IQTT</a></p> <p>The dataset consists of:</p> <ul> <li> <pre>3diqtt-v2-train.h5 (XZ-compressed)</pre> <strong>(Training Dataset)</strong> <ul> <li> <pre>/labeled</pre> <ul> <li> <pre>/questions</pre> format: [10,000 x 4 x 128 x 128 x 3], corresponding to (10k items) x (reference + 3 answers) x (img width) x (img height) x (RGB), np.float32 in range [0,1]</li> <li> <pre>/answers</pre> format: [10,000], corresponding to (10k answers), np.uint8, one of the following three items: [0,1,2]</li> </ul> </li> <li> <pre>/unlabeled</pre> <ul> <li> <pre>/questions</pre> format: [100,000 x 4 x 128 x 128 x 3], corresponding to (100k items) x (reference + 3 answers) x (img width) x (img height) x (RGB), np.float32 in range [0,1]</li> </ul> </li> </ul> </li> <li> <pre>3diqtt-v2-test.h5</pre> <strong>(Test Dataset)</strong> <ul> <li> <pre>/questions</pre> format: [10,000 x 4 x 128 x 128 x 3], corresponding to (10k items) x (reference + 3 answers) x (img width) x (img height) x (RGB), np.float32 in range [0,1].<br> <strong>Important! This is what you have to evaluate yourself on. We have the correct answers but they are not public.</strong></li> </ul> </li> <li> <pre>3diqtt-v2-val.h5</pre> <strong>(Validation Dataset)</strong> <ul> <li> <pre>/questions</pre> format: [10,000 x 4 x 128 x 128 x 3], corresponding to (10k items) x (reference + 3 answers) x (img width) x (img height) x (RGB), np.float32 in range [0,1]</li> <li> <pre>/answers</pre> format [10,000], corresponding to (10k answers), np.uint8, one of the following three items: [0,1,2]</li> </ul> </li> </ul> <p> </p> <p><strong>Important:</strong> Before use, the main training dataset (3diqtt-v2-train.h5.xz) needs to be decompressed. This can take up to 24h depending on your hardware. We apologize for any inconvenience caused by this. The uncompressed file has a size of ~74GB. The reason for this compression was a restriction on the size of individual files. The command for decompression is "<strong>unxz</strong><strong> 3diqtt-v2-train.h5.xz</strong>" on Unix machines.</p> <p><strong>If you use this dataset, please cite it.</strong></p>
Fig. 6.1. Shell digitised with different methods. The photogrammetry model was captured with a 100 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 6.1. Shell digitised with different methods. The photogrammetry model was captured with a 100 mm Macro lens and processed with Agisoft Photoscan. The visual comparison of the mollusc shows a similar level of detail between photogrammetry and MechScan for the external surfaces, with still a bit more detail for the MechScan. The HDI Advance has a much lower resolution.
Fig. 5.6 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 5.6. Decimation of a 3D model. The four parts show a 3D model in various degrees of reducing the amount of faces. In the left upper corner is the original and rotating clockwise are the models at 50%, 75% and 90% decimation. Until 75% there is hardly any difference noticeable, while at 90% the cracks become less deep and the faces become more visible.
Fig. 6.2 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 6.2. Texture comparison between photogrammetry and MechScan (above) and an actual picture captured by a Canon 700D with 100 mm macro lens of the shell below.
Fig. 5.3 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 5.3. Micro-contrast enhancement in DxO OpticsPro 11. A crop of the original image is on the left, one of the post-processed pictures on the right.
Fig. 5.4 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 5.4. Micro-contrast enhancement in DxO OpticsPro 11. The original image is on the left, the postprocessed picture on the right. The post-processed picture looks more crisp and shows more details than the original one as the washed-out appearance has gone.
Fig. 6.14. Ishango rod. The left 3D model was acquired with a in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 6.14. Ishango rod. The left 3D model was acquired with a µCT many years ago. The middle one is scanned with the MechScan structured light scanner. The right one is the combination of both the µCT scan, the structured light scan and the texture of the photogrammetry model.
Fig. 5.1 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 5.1. Relighting in DxO OpticsPro 11. The original image is on the left, the post-processed picture on the right. The underexposed image is now corrected without the need to take new images.
Fig. 4.9 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 4.9. Pre-Columbian handle of an incense shovel from the Royal Museum of Art and History collections. UV fluorescence photogrammetry model. In this case, fluorescence enables to enhance the glue (fluorescing in green). https://sketchfab.com/models/2d82a98be64c48b89cada459b81bd0ab
Fig. 4.7 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 4.7. Enhancing the legibility of a specimen. The picture on the left represents the specimen captured under white light, while the picture on the right displays the specimen under UV light. Part of the reflections is reduced under UV light allowing to display more contrasted structures.
Fig. 4.5 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 4.5. Detail of the Halszkaraptor fossil from Mongolia. In white light on the left, in UV fluorescence on the right. The UV fluorescence image displays restorations of the fossils and treatment applied to preserve it.
Fig. 3.20. 3D in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 3.20. 3D model of a Dorylus ant (size: 1.5 cm) based upon focus stacked images, textured model is on the left, the view of only the mesh is on the right. The VCM option in Agisoft Photoscan is chosen to include small detail in the 3D model. The tibia spurs are clearly marked. https://sketchfab.com/models/da9aa414bfa64caabfe5c552368b16f0
Fig. 3.15 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 3.15. Part of the Cognisys StackShot 3X Deluxe Kit, reassembled for the photogrammetry purpose. The two rotary tables are mounted perpendicular to each other, whereby rotary table A moves a steel angle with rotary table B fixed at the end.
Fig. 3.13 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 3.13. Alnus japonica (2 cm long) scanned with DISC3D. A. EDOF-image. B. 3D-model (vcm) from 807 cameras. C. 3D-model from 398 cameras.
Fig. 3.37 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 3.37. Excavation site scanned with the Gotcha infrared sensor. On the left is the site without texture, on the right with texture. The excavation site pictured measures approximately 4×4 m.
Fig. 3.10 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 3.10. Photogrammetry model of a Costa Rican Sacrificing Warrior (800–1300 AD) in Basalt (RMAH collections). The possibility of viewing the model without the texture has improved the visibility of the belt markings. https://sketchfab.com/models/03a9c7c61cdf48c8845498d1a6b19a73
Fig. 3.9 in Handbook of best practice and standards for 2D+ and 3D imaging of natural history collections
Fig. 3.9. Example of photogrammetry model of Argonauta tuberculata https://sketchfab.com/models/daed659ee685452b91d8f8c91dff761b
ScienceDex guides
Understand access before you commit
These curated guides explain access requirements, typical timelines, costs, and reuse considerations for widely used research datasets.
Allen Brain Atlas
Allen Brain Atlas is an Allen Institute collection of brain map atlases, datasets, APIs, and analysis tools covering mouse, human, and non-human primate brain resources.
Annotated Behaviour and Observability Dataset (ABODe)
ABODe is a University of Edinburgh DataShare dataset for behavior classification in group-housed mice using home-cage video, identities, bounding boxes, ground-plate positions, and annotator labels.
DANDI Archive for NWB datasets
DANDI is a BRAIN Initiative archive for publishing and sharing neurophysiology data, including electrophysiology, optophysiology, and behavioral data packaged as NWB and related standards.
International Brain Laboratory public data
The International Brain Laboratory public data releases expose standardized mouse decision-making experiments, including Neuropixels recordings, widefield calcium imaging, behavior, and session metadata accessed through the ONE API.
OpenNeuro
OpenNeuro is a free, open platform for sharing neuroimaging datasets, with public search, dataset pages, and download paths for web, S3, DataLad, and the OpenNeuro CLI.