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.
75
datasets available to search
ShareScore release 0.9.0
Dataset results
75 results for “pre-training”
Pre-trained models for segmentation and tracking of Coronal Bright Fronts from SDO AIA Base Difference images
<p>Here we present pretrained U-NET-based models followed by SDO AIA Base Difference(BD) validation set after intensity tresholding [-50;150] with predicted feature masks samples. <br>We provide a command-line Python utility for image segmentation using our CNNs designed to process images of solar eruptive phenomena. The https://gitlab.com/iahelio/helios_cnn repository includes regularly updated and newly published models. </p> <p>First model we present is designed to predict the likelihood of each pixel belonging to a certain class or feature in the solar image. A probabilistic output allows for a more nuanced interpretation of ambiguous region. The output can be converted into binary masks through thresholding. The range of values also gives insights into the model's confidence</p> <p>We also present sample segmentation results and the second model designed to produce binary masks.</p>
Pre-training Audio Embeddings
<p>Pre-trained audio embeddings (VGGish, OpenL3, YAMNet) of <a href="https://www.upf.edu/web/mtg/irmas">IRMAS</a> and <a href="https://zenodo.org/record/1432913">OpenMIC-2018</a> datasets, released in the following paper:</p> <p>Changhong Wang, Brian McFee, and Gaël Richard. "<strong>Transfer Learning and Bias Correction with Pre-trained Audio Embeddings</strong>". <em>Proceedings of the <a href="https://ismir2023.ismir.net/">International Society for Music Information Retrieval (ISMIR) Conference</a></em>, 2023.</p>
Pre-training with simulated ultrasound images for breast mass segmentation and classification - dataset
<p>Dataset assosiated with the MICCAI Workshop on Data Engineering in Medical Imaging paper: "Pre-training with Simulated Ultrasound Images for Breast Mass Segmentation and Classification"</p>
Sentence/Table Pair Data from Wikipedia for Pre-training with Distant-Supervision
<p>This is the dataset used for pre-training in "<em>ReasonBERT: Pre-trained to Reason with Distant Supervision</em>", EMNLP'21.</p> <p>There are two files:</p> <p>sentence_pairs_for_pretrain_no_tokenization.tar.gz -> Contain only sentences as evidence, Text-only</p> <p>table_pairs_for_pretrain_no_tokenization.tar.gz -> At least one piece of evidence is a table, Hybrid</p> <p>The data is chunked into multiple tar files for easy loading. We use <a href="https://github.com/webdataset/webdataset">WebDataset</a>, a PyTorch Dataset (IterableDataset) implementation providing efficient sequential/streaming data access.</p> <p>For pre-training code, or if you have any questions, please check our GitHub repo https://github.com/sunlab-osu/ReasonBERT</p> <p>Below is a sample code snippet to load the data</p> <pre><code class="language-python">import webdataset as wds # path to the uncompressed files, should be a directory with a set of tar files url = './sentence_multi_pairs_for_pretrain_no_tokenization/{000000...000763}.tar' dataset = ( wds.Dataset(url) .shuffle(1000) # cache 1000 samples and shuffle .decode() .to_tuple("json") .batched(20) # group every 20 examples into a batch ) # Please see the documentation for WebDataset for more details about how to use it as dataloader for Pytorch # You can also iterate through all examples and dump them with your preferred data format</code></pre> <p>Below we show how the data is organized with two examples.</p> <p>Text-only</p> <pre><code>{'s1_text': 'Sils is a municipality in the comarca of Selva, in Catalonia, Spain.', # query sentence 's1_all_links': { 'Sils,_Girona': [[0, 4]], 'municipality': [[10, 22]], 'Comarques_of_Catalonia': [[30, 37]], 'Selva': [[41, 46]], 'Catalonia': [[51, 60]] }, # list of entities and their mentions in the sentence (start, end location) 'pairs': [ # other sentences that share common entity pair with the query, group by shared entity pairs { 'pair': ['Comarques_of_Catalonia', 'Selva'], # the common entity pair 's1_pair_locs': [[[30, 37]], [[41, 46]]], # mention of the entity pair in the query 's2s': [ # list of other sentences that contain the common entity pair, or evidence { 'md5': '2777e32bddd6ec414f0bc7a0b7fea331', 'text': 'Selva is a coastal comarque (county) in Catalonia, Spain, located between the mountain range known as the Serralada Transversal or Puigsacalm and the Costa Brava (part of the Mediterranean coast). Unusually, it is divided between the provinces of Girona and Barcelona, with Fogars de la Selva being part of Barcelona province and all other municipalities falling inside Girona province. Also unusually, its capital, Santa Coloma de Farners, is no longer among its larger municipalities, with the coastal towns of Blanes and Lloret de Mar having far surpassed it in size.', 's_loc': [0, 27], # in addition to the sentence containing the common entity pair, we also keep its surrounding context. 's_loc' is the start/end location of the actual evidence sentence 'pair_locs': [ # mentions of the entity pair in the evidence [[19, 27]], # mentions of entity 1 [[0, 5], [288, 293]] # mentions of entity 2 ], 'all_links': { 'Selva': [[0, 5], [288, 293]], 'Comarques_of_Catalonia': [[19, 27]], 'Catalonia': [[40, 49]] } } ,...] # there are multiple evidence sentences }, ,...] # there are multiple entity pairs in the query }</code></pre> <p>Hybrid</p> <pre><code>{'s1_text': 'The 2006 Major League Baseball All-Star Game was the 77th playing of the midseason exhibition baseball game between the all-stars of the American League (AL) and National League (NL), the two leagues comprising Major League Baseball.', 's1_all_links': {...}, # same as text-only 'sentence_pairs': [{'pair': ..., 's1_pair_locs': ..., 's2s': [...]}], # same as text-only 'table_pairs': [ 'tid': 'Major_League_Baseball-1', 'text':[ ['World Series Records', 'World Series Records', ...], ['Team', 'Number of Series won', ...], ['St. Louis Cardinals (NL)', '11', ...], ...] # table content, list of rows 'index':[ [[0, 0], [0, 1], ...], [[1, 0], [1, 1], ...], ...] # index of each cell [row_id, col_id]. we keep only a table snippet, but the index here is from the original table. 'value_ranks':[ [0, 0, ...], [0, 0, ...], [0, 10, ...], ...] # if the cell contain numeric value/date, this is its rank ordered from small to large, follow TAPAS 'value_inv_ranks': [], # inverse rank 'all_links':{ 'St._Louis_Cardinals': { '2': [ [[2, 0], [0, 19]], # [[row_id, col_id], [start, end]] ] # list of mentions in the second row, the key is row_id }, 'CARDINAL:11': {'2': [[[2, 1], [0, 2]]], '8': [[[8, 3], [0, 2]]]}, } 'name': '', # table name, if exists 'pairs': { 'pair': ['American_League', 'National_League'], 's1_pair_locs': [[[137, 152]], [[162, 177]]], # mention in the query 'table_pair_locs': { '17': [ # mention of entity pair in row 17 [ [[17, 0], [3, 18]], [[17, 1], [3, 18]], [[17, 2], [3, 18]], [[17, 3], [3, 18]] ], # mention of the first entity [ [[17, 0], [21, 36]], [[17, 1], [21, 36]], ] # mention of the second entity ] } } ] }</code></pre> <p> </p> <p> </p> <p> </p>
Deep Reference Mining from Scholarly Literature in the Arts and Humanities - Pre-trained word embeddings
<p>Pre-trained word vectors of dimensionality 100 and 300 for the publication: Deep Reference Mining from Scholarly Literature in the Arts and Humanities, submitted to Frontiers in Digital Humanities.</p> <p>The corpus of scholarly publications from which these vectors were trained is under copyright, therefore we publish these vectors for reproducibility. Please refer to the publication's repository for further details: <a href="https://github.com/dhlab-epfl/LinkedBooksDeepReferenceParsing">https://github.com/dhlab-epfl/LinkedBooksDeepReferenceParsing</a>.</p> <p>These vectors were trained using Gensim 3.1.0. The corpus was preprocessed as follows:</p> <ol> <li>word tokenization with NLTK word_punct tokenizer.</li> <li>digits were converted into the $NUM$ token</li> <li>words less frequent than 5 times, for every document, were converted to the $UNK$ token</li> <li>vectors were trained using the function: Word2Vec(window=5, min_count=5, sg=1)</li> </ol>
Ecore Metamodels and EcoreBERT Pre-trained Language Model
<p>This dataset contains ecore metamodels from the MAR dataset transformed into tree representations. The original dataset can be found here: <a href="http://mar-search.org/experiments/models20/">http://mar-search.org/experiments/models20/</a></p> <p>The data contained in this repository were used to conduct the experiments in the paper: <strong>Recommending Metamodel Concepts during Modeling Activities with Pre-Trained Language Models. </strong>Link to the paper: <a href="https://arxiv.org/abs/2104.01642">https://arxiv.org/abs/2104.01642</a></p> <p>The data are organized as follows:</p> <ul> <li>model : our model trained on the tree representations of metamodels with RoBERTa architecture.</li> <li>tokenizers : the byte-level BPE tokenizer we used to train our model.</li> <li>train : the training data separated into a training and validation set.</li> <li>test : the test data of all experiments conducted in the paper.</li> </ul> <p>This data repository is linked with the following Github repository containing our code: <a href="https://github.com/mweyssow/ecore-bert">https://github.com/martiwey/metamodel-concepts-bert</a></p>
MS2Query pre-trained datasets needed for library matching
<p>The models, embeddings, sqlite file with metadata and classifiers identifiers needed for running MS2Query (https://github.com/iomega/ms2query)</p>
BioVAE: a pre-trained latent variable language model for biomedical text mining
<p>We release BioVAE, the first large-scale pre-trained latent variable language model for the biomedical domain, which uses the OPTIMUS framework to train on large volumes of biomedical text.</p> <p>This version contains the pre-trained models for text mining tasks such as named entity recognition or relation extraction, and text generation task.</p> <p>Explanation of each file: (lt32: latent_size = 32, beta05: beta=0.5)</p> <ul> <li>pm-full-lt32-beta00</li> <li>pm-full-lt32-beta05</li> <li>pm-full-lt768-beta00</li> <li>pm-full-lt768-beta05</li> <li>pm-full-generation</li> </ul>
A novel strategy for fully automated segmentation of supratentorial meningiomas: Use of pre-trained models and inclusion of normal brain images
<p>This repository is accompanying MRI datasets under the journal, titled: <strong>A novel strategy for fully automated segmentation of supratentorial meningiomas: Use of pre-trained models and inclusion of normal brain images</strong>. </p> <p>Nii_data.tar.gz (zipped) file includes MRI images of all patients described in the paper that are formatted as .nii.</p>
dMRI-RCNN Pre-Trained Weights - 3D 10in xmax
<p>Pre-trained weights for <a href="http://github.com/m-lyon/dMRI-RCNN">dMRI-RCNN</a>. 3D RCNN model, for q_in = 10, with varying xmax normalisation.</p>
BiodivBERT: Pre-training Corpora DOIs
<p>This repository contains the DOIs we used to construct the pre-training corpora for BiodivBERT model.</p> <p>BiodivBert<sub>Abs<sup> </sup></sub>uses the abstracts DOIs while BiodivBERT<sub>Abs+Full</sub> uses both of them. </p>
Training and validation data used to produce the pre-trained model for the TomoTwin paper.
<p>This datasets represents the training and validation data that was used to produce the pre-trained model for the TomoTwin paper. Please see 10.5281/zenodo.6637357 for the raw tomograms.</p>
Semi-Supervised Pre-trained Foundation Model for 3D Structural Feature Analysis of Seismic Images
<p>Codes, trained model, and datasets for the paper "Semi-Supervised Pre-trained Foundation Model for 3D Structural Feature Analysis of Seismic Images".</p>
An Empirical Comparison of Pre-Trained Models of Source Code
<p>The replication package of the paper "An Empirical Comparison of Pre-Trained Models of Source Code". For the source code, please refer to <a href="https://github.com/NougatCA/FineTuner">https://github.com/NougatCA/FineTuner</a>.</p>
UniFMIR: Pre-training a Foundation Model for Universal Fluorescence Microscopy Image Restoration
<p>This repository contains the preprocessed dataset for [UniFMIR](https://github.com/cxm12/UNiFMIR/). All training and test data involved in the experiments are publicly available datasets. Licenses of the original dataset are applied. You can refer to the Github repository for details.</p> <p>* The 3D denoising/isotropic reconstruction/projection datasets can be downloaded from [Content Aware Image Restoration dataset](https://publications.mpi-cbg.de/publications-sites/7207/). `Projection_Flywing/train_data/my_training_data.npz` are generated according to the [CSBDeep](http://csbdeep.bioimagecomputing.com/doc/).</p> <p>* The SR dataset can be downloaded from [BioSR dataset](https://doi.org/10.6084/m9.figshare.13264793). The dataset is augmented according to the instructions in [DFCAN](https://github.com/qc17-THU/DL-SR/tree/main#train-a-new-model) and `my_training_data.npz` files are generated following [CSBDeep](http://csbdeep.bioimagecomputing.com/doc/datagen.html). </p> <p>* The Volumetric reconstruction dataset are from [VCD-LFM dataset](https://doi.org/10.5281/zenodo.4390067). The dataset is prepared according to the instructions in [VCD-Net](https://github.com/feilab-hust/VCD-Net).</p> <p>* DeepBacs dataset can be downloaded from [DeepBacs dataset](https://zenodo.org/record/6460867). We split the dataset into 5 folds for cross-validation. Shareloc dataset can be downloaded from [Shareloc dataset](https://zenodo.org/record/7234161).</p> <p> </p> <p>The data paths should be as follows:</p> <p>```</p> <p>VCD/vcdnet/</p> <p>CSB/DataSet/</p> <p> Denoising_Planaria/</p> <p> Denoising_Tribolium/</p> <p> Isotropic/Isotropic_Liver/</p> <p> Projection_Flywing/</p> <p> BioSR_WF_to_SIM/DL-SR-main/dataset/</p> <p> Synthetic_tubulin_gfp/</p> <p> Synthetic_tubulin_granules/</p> <p>DeepBacs/</p> <p>Shareloc/</p> <p>```</p>
Phenanthrene: TD-DFTB datasets, pre-trained SchNet models and initial coniditions for TSH
<p><em>Data associated with the paper entitled </em></p> <p><strong>On application of Deep Learning to simplified quantum-classical dynamics in electronically excited states</strong></p> <ol> <li>Three TD-DFTB datasets (<strong>sX_10_force.db</strong>) have been produced using the <a href="https://wiki.fysik.dtu.dk/ase/">Atomic Simulation Environment</a> (ASE) coupled to <a href="http://demon-nano.ups-tlse.fr/">deMon-Nano</a> code for the linear response Time-Dependent Density Functional based Tight-Binding (TD-DFTB) calculations. Each dataset contains 10000 TD-DFTB electronic structure calculations for a given excited singlet state (S<sub>2</sub>/S<sub>3</sub>/S<sub>4</sub>) of a neutral phenanthrene molecule. Each database entry contains Cartesian atomic coordinates as well as potential energy and atomic forces for a given excited state at a given geometry. Since ASE has been used, all physical quantities are stored in the corresponding units (e.g. eV for energy or eV/Å for forces). The file format is SQLite as provided by the ASE;</li> <li>Three pre-trained Deep Learning models (<strong>best_model_sX</strong>) for a given excited singlet state have been produced using <a href="https://schnetpack.readthedocs.io/en/stable/">SchNetPack</a> package, which implements the SchNet architecture for atomistic simulations. Each model has been trained using the corresponding TD-DFTB dataset from #1. The file format is binary as provided by the SchNetPack;</li> <li><a href="https://zenodo.org/api/files/f1925cb5-66a8-4c6f-809b-3414f0cbc1d5/500_init_conditions.tar.gz"><strong>500_init_conditions.tar.gz</strong> </a> contains 500 initial conditions (Cartesian coordinates and velocities), which can be used for Trajectory Surface Hopping (TSH) simulations with or without the pre-trained models from #2.</li> </ol> <p> </p>
Pre-trained word2vec models for ``Easy over Hard: A Case Study on Deep Learning''
<p>Since the whole stack overflow dump is so big, we can't easily handle well. Here, we provide 10 pre trained word2vec models with different seeds.</p> <p> </p> <p>More details about how to use it, please see paper </p>
ANNs pre-trained on Retinal Waves
<p>Different Artificial Neural Networks (saved weights), some only pre-trained either on rwave-1024 or rwave-4096 or FractalDB1000 datasets; some fine-tuned or trained from scratch (pt_none_ft... or pt__ft... or ...scratch...) on CIFAR10/100 or ImageNet1k.</p><p>Retinal Waves for Pre-Training Artificial Neural Networks Mimicking Real Prenatal Development - see https://github.com/BennyCa/ReWaRD for filter visualization and further fine-tuning possibilities</p><p>Pre-training and fine-tuning was conducted using the codebase https://github.com/hirokatsukataoka16/FractalDB-Pretrained-ResNet-PyTorch</p>
MOST-GAN Pre-trained Model
<p><strong>Introduction</strong></p> <p>Recent advances in generative adversarial networks (GANs) have led to remarkable achievements in face image synthesis. While methods that use style-based GANs can generate strikingly photorealistic face images, it is often difficult to control the characteristics of the generated faces in a meaningful and disentangled way. Prior approaches aim to achieve such semantic control and disentanglement within the latent space of a previously trained GAN. In contrast, we propose a framework that a priori models physical attributes of the face such as 3D shape, albedo, pose, and lighting explicitly, thus providing disentanglement by design. Our method, MOST-GAN, integrates the expressive power and photorealism of style-based GANs with the physical disentanglement and flexibility of nonlinear 3D morphable models, which we couple with a state-of-the-art 2D hair manipulation network. MOST-GAN achieves photorealistic manipulation of portrait images with fully disentangled 3D control over their physical attributes, enabling extreme manipulation of lighting, facial expression, and pose variations up to full profile view. </p> <p>To foster further research into this topic, we are publicly releasing our pre-trained model for MOST-GAN. Please see our AAAI paper titled [MOST-GAN: 3D Morphable StyleGAN for Disentangled Face Image Manipulation](https://arxiv.org/abs/2111.01048) for details.</p> <p><strong>At a Glance</strong></p> <p>-The size of the unzipped model is ~300MB.</p> <p>-The unzipped folder contains: (i) a README.md file and (ii) ./checkpoints/checkpoint01.pt pre-trained model. The pre-trained model could be loaded in our publicly released MOST-GAN implementation.</p> <p><strong>Citation</strong></p> <p>If you use the MOST-GAN data in your research, please cite our paper:</p> <pre><code>@inproceedings{medin2022most, title={MOST-GAN: 3D morphable StyleGAN for disentangled face image manipulation}, author={Medin, Safa C and Egger, Bernhard and Cherian, Anoop and Wang, Ye and Tenenbaum, Joshua B and Liu, Xiaoming and Marks, Tim K}, booktitle={Proceedings of the AAAI conference on artificial intelligence}, volume={36}, number={2}, pages={1962--1971}, year={2022} } </code></pre> <p><strong>License</strong></p> <p>The MOST-GAN data is released under <a href="https://creativecommons.org/licenses/by-sa/4.0/">CC-BY-SA-4.0 license</a>.</p> <p>All data:</p> <pre><code>Created by Mitsubishi Electric Research Laboratories (MERL), 2022,2023 SPDX-License-Identifier: CC-BY-SA-4.0 </code></pre> <p> </p>
LUVLi Pre-trained Model
<p><strong>Introduction</strong></p> <p>Modern face alignment methods have become quite accurate at predicting the locations of facial landmarks, but they do not typically estimate the uncertainty of their predicted locations nor predict whether landmarks are visible. In this paper, we present a novel framework for jointly predicting landmark locations, associated uncertainties of these predicted locations, and landmark visibilities. We model these as mixed random variables and estimate them using a deep network trained with our proposed Location, Uncertainty, and Visibility Likelihood (LUVLi) loss. In addition, we release an entirely new labeling of a large face alignment dataset with over 19,000 face images in a full range of head poses. Each face is manually labeled with the ground-truth locations of 68 landmarks, with the additional information of whether each landmark is unoccluded, self-occluded (due to extreme head poses), or externally occluded. Not only does our joint estimation yield accurate estimates of the uncertainty of predicted landmark locations, but it also yields state-of-the-art estimates for the landmark locations themselves on multiple standard face alignment datasets. Our method’s estimates of the uncertainty of predicted landmark locations could be used to automatically identify input images on which face alignment fails, which can be critical for downstream tasks.</p> <p>To foster further research into this topic, we are publicly releasing our pre-trained LUVLi models. Please see our CVPR 2020 paper titled <a href="https://arxiv.org/abs/2004.02980">LUVLi Face Alignment: Estimating Landmarks’ Location, Uncertainty, and Visibility Likelihood</a> for details</p> <p><strong>At a Glance</strong></p> <p>-The size of the unzipped model is ~700MB.</p> <p>-The unzipped folder contains: (i) a README.md file and (ii) pre-trained models and logs. The pre-trained models could be loaded in our publicly released LUVLi implementation.</p> <p><strong>Other Resources</strong></p> <p><strong>Citation</strong></p> <p>If you use the LUVLi data in your research, please cite our paper:</p> <pre><code>@inproceedings{kumar2020luvli, title={{LUVLi} Face Alignment: Estimating Landmarks' Location, Uncertainty, and Visibility Likelihood}, author={Kumar, Abhinav and Marks, Tim K. and Mou, Wenxuan and Wang, Ye and Jones, Michael and Cherian, Anoop and Koike-Akino, Toshiaki and Liu, Xiaoming and Feng, Chen}, booktitle={IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, year={2020} } </code></pre> <p><strong>License</strong></p> <p>The LUVLi data is released under <a href="https://creativecommons.org/licenses/by-sa/4.0/">CC-BY-SA-4.0 license</a>.</p> <p>All data:</p> <pre><code>Created by Mitsubishi Electric Research Laboratories (MERL), 2022,2023 SPDX-License-Identifier: CC-BY-SA-4.0 </code></pre>
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.