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.
187
datasets available to search
ShareScore release 0.9.0
Dataset results
187 results for “Training Set”
QLKNN11D training set
<p><strong>QLKNN11D training set</strong></p> <p>This dataset contains a large-scale run of ~1 billion flux calculations of the quasilinear gyrokinetic transport model QuaLiKiz. QuaLiKiz is applied in numerous tokamak integrated modelling suites, and is openly available at <a href="https://gitlab.com/qualikiz-group/QuaLiKiz/">https://gitlab.com/qualikiz-group/QuaLiKiz/</a>. This dataset was generated with the 'QLKNN11D-hyper' tag of QuaLiKiz, equivalent to 2.8.1 apart from the negative magnetic shear filter being disabled. See <a href="https://gitlab.com/qualikiz-group/QuaLiKiz/-/tags/QLKNN11D-hyper">https://gitlab.com/qualikiz-group/QuaLiKiz/-/tags/QLKNN11D-hyper</a> for the in-repository tag.</p> <p>The dataset is appropriate for the training of learned surrogates of QuaLiKiz, e.g. with neural networks. See <a href="https://doi.org/10.1063/1.5134126">https://doi.org/10.1063/1.5134126</a> for a Physics of Plasmas publication illustrating the development of a learned surrogate (QLKNN10D-hyper) of an older version of QuaLiKiz (2.4.0) with a 300 million point 10D dataset. The paper is also available on arXiv <a href="https://arxiv.org/abs/1911.05617">https://arxiv.org/abs/1911.05617</a> and the older dataset on Zenodo <a href="https://doi.org/10.5281/zenodo.3497066">https://doi.org/10.5281/zenodo.3497066</a>. For an application example, see Van Mulders et al 2021<a href="http://https://doi.org/10.1088/1741-4326/ac0d12"> https://doi.org/10.1088/1741-4326/ac0d12</a>, where QLKNN10D-hyper was applied for ITER hybrid scenario optimization. For any learned surrogates developed for QLKNN11D, the effective addition of the alphaMHD input dimension through rescaling the input magnetic shear (s) by s = s - alpha_MHD/2, as carried out in Van Mulders et al., is recommended.</p> <p>Related repositories:</p> <ul> <li>General QuaLiKiz documentation <a href="https://qualikiz.com">https://qualikiz.com</a></li> <li>QuaLiKiz/QLKNN input/output variables naming scheme <a href="https://qualikiz.com/QuaLiKiz/Input-and-output-variables">https://qualikiz.com/QuaLiKiz/Input-and-output-variables</a></li> <li>Training, plotting, filtering, and auxiliary tools <a href="https://gitlab.com/Karel-van-de-Plassche/QLKNN-develop">https://gitlab.com/Karel-van-de-Plassche/QLKNN-develop</a></li> <li>QuaLiKiz related tools <a href="https://gitlab.com/qualikiz-group/QuaLiKiz-pythontools">https://gitlab.com/qualikiz-group/QuaLiKiz-pythontools</a></li> <li>FORTRAN QLKNN implementation with wrapper for Python and MATLAB <a href="https://gitlab.com/qualikiz-group/QLKNN-fortran">https://gitlab.com/qualikiz-group/QLKNN-fortran</a></li> <li>Weights and biases of 'hyperrectangle style' QLKNN <a href="https://gitlab.com/qualikiz-group/qlknn-hype">https://gitlab.com/qualikiz-group/qlknn-hype</a></li> </ul> <p><strong>Data exploration</strong></p> <p>The data is provided in 43 netCDF files. We advise opening single datasets using <a href="https://xarray.dev/">xarray </a>or multiple datasets out-of-core using <a href="https://www.dask.org/">dask</a>. For reference, we give the load times and sizes of a single variable that just depends on the scan size `dimx` below. This was tested single-core on a Intel Xeon 8160 CPU at 2.1 GHz and 192 GB of DDR4 RAM. Note that during loading, more memory is needed than the final number.</p> <table align="center"> <caption>Timing of dataset loading</caption> <thead> <tr> <th scope="col">Amount of datasets</th> <th scope="col">Final in-RAM memory (GiB)</th> <th scope="col"> <p>Loading time single var</p> (M:SS)</th> </tr> </thead> <tbody> <tr> <td>1</td> <td>10.3</td> <td>0:09</td> </tr> <tr> <td>5</td> <td>43.9</td> <td>1:00</td> </tr> <tr> <td>10</td> <td>63.2</td> <td>2:01</td> </tr> <tr> <td>16</td> <td>98.0</td> <td>3:25</td> </tr> <tr> <td>17</td> <td>Out Of Memory</td> <td>x:xx</td> </tr> </tbody> </table> <p><strong>Full dataset</strong></p> <p>The full dataset of QuaLiKiz in-and-output data is available on request. Note that this is 2.2 TiB of netCDF files!</p>
PANDEM-2 European COVID-19 training data set
<p>The PANDEM-2 COVID-19 European training dataset is a large collection of time series either of real or realistic synthetic (generated) data and indicators associated with the European pandemic response to the COVID-19 pandemic. It is intended to be used for training in pandemic management.</p> <p> </p> <p>This dataset is the result of a data gathering requirement process for pandemic management involving feedback and inputs from several public health and first responder professionals as well as researchers and military personnel directly involved in the European COVID-19 pandemic response. This work is part of the PANDEM-2 project funded by the <em>Horizon 2020 Secure Societies</em> program. To collect this data, an open source software was developed named PANDEM-Source allowing reproducibility and customisation of this dataset. </p> <p> </p> <p>The dataset includes indicators for cases, deaths, hospitalisation, testing and laboratory data including pathogen genomic information, vaccination, non-pharmaceutical interventions, participatory surveillance, social media, flights resources (human and material such as beds or vaccines), and contact tracing activities. When no open available data was found, realistic synthetic data and indicators were generated with the goal of producing a data set to be used for pandemic management training. </p> <p>The project received funding from the European Union’s Horizon 2020 Research and Innovation programme under the Grant Agreement No. 883285. The material presented and views expressed here are the responsibility of the author(s) only. The EU Commission takes no responsibility for any use made of the information set out.<br> References</p> <p> </p> <p> </p>
RDP Classifier 2.14 and the RDP bacterial and archaeal taxonomy training set No. 19
<p>RDP Classifier 2.14 (August 2023) Release Note:</p><p>The Bacteria and Archaea hierarchy model used by RDP Classifier has been updated to training set No. 19. The new version has over 600 genera and 2500 species added since last version No. 18 released in July 2020. The information that is used to update the RDP taxonomy to training set version No. 19, and RDP Classifier version 2.14 came from publicly available scientific articles and public sequence repository, mostly from International Journal of Systematic and Evolutionary Microbiology (IJSEM), the All-Species Living Tree Project (LTP) and GenBank. </p><p>It is worth noting that most of the phyla have new names, according to article "</p><p>Oren A, Garrity GM. Valid publication of the names of forty-two phyla of prokaryotes. Int J Syst Evol Microbiol. 2021 Oct;71(10). doi: 10.1099/ijsem.0.005056. PMID: 34694987."</p><p>In addition to the files to train and run the RDP Classifier, new file formats are made available to accommodate the needs of users:</p><p>1. A new file trainset19_072023_speciesrank.fa has been added to the release in RDPClassifier_16S_trainsetNo19_rawtrainingdata.zip. This file is NOT needed to train the classifier. In addition to sequences, it contains genus, species, strain, type status and taxonomy rank, which are useful for closest species identification using third-party tools (e.g. BLAST).</p><p>2. Two new files in RDPClassifier_16S_trainsetNo19_QiimeFormat.zip to retrain the RDP Classifier included in Qiime2 package.</p>
BTSbot v10 training set
<p>This is the production version of the BTSbot training set, limited to public (programid=1) ZTF alerts. BTSbot is a multi-modal convolutional neural network designed for real-time identification bright extragalactic transients in Zwicky Transient Facility (ZTF) data. BTSbot provides a bright transient score to individual ZTF detections using their image data and 25 extracted features. BTSbot is able to eliminate the need for daily visual inspection of new transients by automatically identifying and requesting spectroscopic follow-up observations of new bright transient candidates. </p> <p>The training data is split into two zipped files. metadata_v10.zip contains alert packet features for the alerts in the train, validation, and test splits stored a separate .csv files. images_v10.zip contains the corresponding image cutouts stored as three .npy files. The BTSbot source code contains routines for reading these files and training a model on them. They can also easily be loaded with pandas.read_csv() and numpy.load().</p> <p>This training set data is necessary for reproducing the results of the BTSbot study, although this dataset only contains ZTF public data while the production BTSbot model also trained on ZTF partnership data.</p> <p>If you use reference this data or BTSbot please cite the <a href="https://ui.adsabs.harvard.edu/abs/2024arXiv240115167R/abstract" target="_blank" rel="noopener">BTSbot paper</a>.</p> <p> </p>
Gravity Spy Training Set
<p>We present an updated version of the training set used in the Gravity Spy citizen science project. This training set, curated further from the one discussed in detail <a href="https://www.sciencedirect.com/science/article/pii/S0020025518301634">here</a> and available here <a href="https://doi.org/10.5281/zenodo.1476156">10.5281/zenodo.1476156</a>, was utilized to train the convolutional neural network employed in the Gravity Spy project. We anticipate moving forward to release more labelled Gravity Spy data sets, including a data set containing the annotations provided by our citizen science volunteers.</p> <p><strong>Data Set Information</strong></p> <p>There are three files provided in this data set</p> <ul> <li><strong>trainingset_v1d1_metadata.csv</strong> <ul> <li>This file has many columns, <em>gravityspy_id, label, </em>and <em>sample_type.</em><em> gravityspy_id </em>is the unique 10 character hash given to every Gravity Spy sample. <em>label</em> is the string label of the sample. <em>sample_type </em>indicates whether this sample was used in the paper for testing training or validating the models. This is provided for those who would like to do direct comparisons to the network described in the paper.</li> <li>Additional columns contain some metadata information about the glitch <ul> <li><em>event_time,ifo,peak_time,peak_time_ns,start_time,start_time_ns,duration,search,process_id,event_id,peak_frequency,central_freq,bandwidth,channel,amplitude,snr,confidence,chisq,chisq_dof,param_one_name,param_one_value,</em></li> </ul> </li> </ul> </li> <li><em><strong>trainingsetv1d1.h5</strong></em> <ul> <li>This file contains the exact arrays used in the paper for every Gravity Spy sample. Each Gravity Spy sample is defined by four different images with varying temporal duration, <em>0.5, 1.0, 2.0, and 4.0</em> second, respectively. This also determines the naming conventions of the PNGs: <em>interferometer_gravityspyid_spectrogram_duration.png (e.g. H1_Fv3p6eROvA_spectrogram_0.5.png, H1_Fv3p6eROvA_spectrogram_1.0.png, H1_Fv3p6eROvA_spectrogram_2.0.png, H1_Fv3p6eROvA_spectrogram_4.0.png</em>).</li> <li>This file contains all the information needed for each sample in the Gravity Spy dataset (i.e. the label, the sample type of the sample, the unique id of the sample, and the image data for that sample. <ul> <li>/1080Lines/validation/xUEyaWr34c Group<br> /1080Lines/validation/xUEyaWr34c/0.5.png Dataset {1, 140, 170}<br> /1080Lines/validation/xUEyaWr34c/1.0.png Dataset {1, 140, 170}<br> /1080Lines/validation/xUEyaWr34c/2.0.png Dataset {1, 140, 170}<br> /1080Lines/validation/xUEyaWr34c/4.0.png Dataset {1, 140, 170}</li> </ul> </li> </ul> </li> <li><strong>trainingsetv1d1.tar.gz</strong> <ul> <li>Contains the raw PNGs of the Gravity Spy training set.</li> <li>The structure of the folder is <em>/"label"/"sample_type"/"pngs"</em></li> </ul> </li> </ul> <p><strong>Data Set Parsing Information</strong></p> <p>To read and crop out the plot axis and labels of the provided PNGs, the following small python code using scikit-image should work.</p> <p>from skimage import io</p> <p>image_data = io.imread("filename_of_image")</p> <p>x=[66, 532]; y=[105, 671]</p> <p>image_data = image_data[x[0]:x[1], y[0]:y[1], :3]</p>
ShiftCrypt training data set for training ConforMine
<p>Data set containing the ShiftCrypt values of the proteins used for training ConforMine. This set is not to be confused the the Molecular Dynamics (MD) data set, also used for training of this model. </p> <p>The data set also contains a python script which recreates the filtering of sequences performed in the training steps of ConforMine, which discards all proteins for which no valid ShiftCrypt predictions were obtained. </p>
QLKNN7D-edge training set
<p><strong>QLKNN7D-edge training set</strong></p> <p>This dataset contains a large-scale run of ~15 million flux calculations of the quasilinear gyrokinetic transport model QuaLiKiz. The dataset is in a parameter regime typical of the L-mode near edge (pedestal forming region). QuaLiKiz is applied in numerous tokamak integrated modelling suites, and is openly available at <a href="https://gitlab.com/qualikiz-group/QuaLiKiz/">https://gitlab.com/qualikiz-group/QuaLiKiz/</a>. This dataset was generated with QuaLiKiz 2.8.4, which includes numerical improvements increasing the robustness of strongly driven (high gradient) calculations typical of the L-mode near-edge. See <a href="https://gitlab.com/qualikiz-group/QuaLiKiz/-/tags/2.8.4">https://gitlab.com/qualikiz-group/QuaLiKiz/-/tags/2.8.4</a> for the in-repository tag.</p> <p>The dataset is appropriate for the training of learned surrogates of QuaLiKiz, e.g. with neural networks. See <a href="https://doi.org/10.1063/1.5134126">https://doi.org/10.1063/1.5134126</a> for a Physics of Plasmas publication illustrating the development of a learned surrogate (QLKNN10D-hyper) of an older version of QuaLiKiz (2.4.0) with a 300 million point 10D dataset. The paper is also available on <a href="https://arxiv.org/abs/1911.05617">arXiv</a> and the older dataset on <a href="https://doi.org/10.5281/zenodo.3497066">Zenodo</a>. For an application example, see <a href="http://https://doi.org/10.1088/1741-4326/ac0d12">Van Mulders et al 2021</a>, where QLKNN10D-hyper was applied for ITER hybrid scenario optimization. An additional, larger, QuaLiKiz dataset is found at <a href="https://zenodo.org/record/8017522">https://zenodo.org/record/8017522</a>. Neither the QLKNN10D or QLKNN11D datasets include L-mode near-edge parameters. For any learned surrogates developed for QLKNN7D-edge, the effective addition of the alphaMHD input dimension through rescaling the input magnetic shear (s) by s = s - alpha_MHD/2, as carried out in Van Mulders et al., is recommended.</p> <p>Related repositories:</p> <ul> <li><a href="https://qualikiz.com">General QuaLiKiz documentation </a></li> <li><a href="https://qualikiz.com/QuaLiKiz/Input-and-output-variables">QuaLiKiz/QLKNN input/output variables naming scheme </a></li> <li><a href="https://gitlab.com/Karel-van-de-Plassche/QLKNN-develop">Training, plotting, filtering, and auxiliary tools </a></li> <li><a href="https://gitlab.com/qualikiz-group/QuaLiKiz-pythontools">QuaLiKiz related tools </a></li> <li><a href="https://gitlab.com/qualikiz-group/QLKNN-fortran">FORTRAN QLKNN implementation with wrapper for Python and MATLAB </a></li> <li><a href="https://gitlab.com/qualikiz-group/qlknn-hyper">Weights and biases of 'hyperrectangle style' QLKNN </a></li> </ul> <p> </p>
Train and Evaluation Code, Road Classification Models and Test set of the paper "Impact of Image Resolution and Image Overlap on the Prediction Performance of Convolutional Neural Networks Trained for Road Classification"
<p>This repository contains the Python scripts built for training and evaluation of the implementation, together with the test data and the resulting road classification models corresponding to the paper "Impact of Image Resolution and Image Overlap on the Prediction Performance of Convolutional Neural Networks Trained for Road Classification". The scripts make use of the Tensorflow with Keras framework and the additional required dependencies.</p> <p>The training and validation set is based on the binary SROADEX dataset (https://zenodo.org/records/6482346) that was re-split into tiles that feature the image resolutions (256 x 256, 512 x 512, and 1024 x 1024 pixels) and image overlaps (0% and 12.5%) considered in this study. The data have been generated using scripts developed in Python using Open Source libraries (GDAL/OGR and MapScript) for rasterization of vector cartography that represents the axes of the different types of roads (urban, interurban and rural). This binary road data contains information from 16 full orthoimages (28.5 km * 18.5 km) with spatial resolution of 0.5 m/pixel from the insular and peninsular Spanish territory. Due to the size on disk of approximately 546 gigabytes, this training and validation data is only available upon request from the corresponding author. The test set has been generated from a novel area of 28.5 km * 18.5 km and features binary road labels. The test sets are provided in the repository for each resolution (with no overlap), so that additional DL models can be evaluated on the same data and compared with the results achieved in this study.</p> <p>The structure of the information shared in this repository is as follows:<br>The scripts have been grouped by tile resolution (256, 512 and 1024). First, the test set and the evaluation script can be found. For each tile resolution, there are two subfolders (corresponding to the "no overlap" and "12.5% overlap"). In each case, the Python scripts for training the models in the three repetitions are shared, and the trained models (H5 format) are shared in compressed form. Finally, for each resolution we also share the testing dataset which consists of two folders.</p> <p>The material is distributed under a CC-BY 4.0 license.</p>
Annotated images from yeast cell lifespans - Training & Test sets - DetecDiv (id01)
<p>This dataset has been generated by manual annotation from timelapse images of yeast cells dividing using the DetecDiv software (see below).</p> <p>It contains ~250 000 images from 250 cellular lifespans (each lifespan is made of between 700 and 1000 images). Each image is classified between 6 classes: "1. unbudded", "2. small", "3. large", "4. dead", "5. empty", "6. clog", according to the subfolder of the image.</p> <p>Besides, this folder also contains a .mat file containing 250 timeseries of classes corresponding to the lifespan of the 200 cells.</p> <p> </p> <p>The dataset used for training (200 cellular lifespans) is in the folder /trainingset, while the dataset used for testing (50cellular lifespans) is in the folder /testset</p> <p>It has been used to train the network <a href="https://doi.org/10.5281/zenodo.5553862">doi.org/10.5281/zenodo.5553862</a> from the software DetecDiv: <a href="https://github.com/gcharvin/DetecDiv">github.com/gcharvin/DetecDiv</a></p> <p><a href="https://biorxiv.org/content/10.1101/2021.10.05.463175v1">biorxiv.org/content/10.1101/2021.10.05.463175v1</a></p> <p> </p> <p><strong>Data type</strong>: 3D microscopy images (3 stacks brightfield) (.tif) + annotation (.mat)</p> <p><strong>Microscopy data type</strong>: Brightfield images with 3 stacks, each stack representing a color of RGB.</p> <p><strong>Imaging</strong>: 20x 0.45 NA brightfield, 6.5µm*6.5µm sCMOS</p> <p><strong>Cell type</strong>: Budding yeast wild type cell (BY4742)</p> <p><strong>File format</strong>: .tif (16-bit RGB, 1 color per z-stack) + .mat</p> <p><strong>Image size</strong>: 60x60x1 (Pixel size: x,y: 325 nm, 3*z: 3*1325 nm)</p> <p> </p> <p><strong>Author(s)</strong>: Théo, ASPERT</p> <p><strong>Contact email</strong>: theo.aspert@gmail.com</p> <p><strong>Affiliation</strong>: IGBMC, Université de Strasbourg</p> <p><strong>Funding bodies</strong>: This work was supported by the Agence Nationale pour la Recherche, the grant ANR-10-LABX-0030-INRT, a French State fund managed by the Agence Nationale de la Recherche under the frame program Investissements d'Avenir ANR-10-IDEX-0002-02.</p>
Data set - Practical teacher-training program on STEAM activity planning
<p>Data set of answers of 14 Brazilian teachers who took part in a practical teacher-training program on STEAM activity planning. </p>
In silico data-set used to train the PGNNIV
<p>The folder contains the data used to train PGNNIVs to unravel the go or grow behaviour of glioblastoma under 4 different parametric models. These data consist on the solutions (in time and space) of eleven simulations with different oxygen boundary conditions.</p> <p>Each folder is named as DATA_"ModelName", where "ModelName" can be: "Sigmoid", "ReLU", "MichaelisMenten" or "Heaviside". Inside each folder, the multidimensional arrays for input and output data for the network training can be found. These arrays have dimension [nExp,TimeStep,x,field], where:</p> <ul> <li>nExp = 11 and corresponds to the number of different configurations or experiments simulated.</li> <li>TimeStep = 1000 and correspond to the different temporal frames where the solution is given.</li> <li>x = 51 and corresponds to the different spatial points where the solution is given.</li> <li>field = 2 and correspond to the different solution fields (1: cells, 2: oxygen).</li> </ul>
Replication package for "To What Extent do Deep Learning-based Code Recommenders Generate Predictions by Cloning Code from the Training Set?
<p>Replication package for "To What Extent do Deep Learning-based Code Recommenders Generate Predictions by Cloning Code from the Training Set?"</p>
MItosis DOmain Generalization Challenge 2022 (MICCAI MIDOG 2022), Training data set (PNG version)
<p>This is the training dataset of the MItosis DOmain Generalization (MIDOG) challenge 2022, held in conjunction with MICCAI 2022. Please find the structured challenge description at 10.5281/zenodo.6362337.</p> <p>The training set consists of 405 tumor cases in total across six tumor types:</p> <ul> <li>Canine Lung Cancer (44 cases, scanned with 3DHistech Pannoramic Scan II)</li> <li>Human Breast Cancer (150 cases, scanned using three scanners, part of MIDOG2021 dataset)</li> <li>Canine Lymphoma (55 cases, scanned with 3DHistech Pannoramic Scan II)</li> <li>Human neuroendocrine tumor (55 cases, scanned with Hamamatsu NanoZoomer XR)</li> <li>Canine Cutaneous Mast Cell Tumor (50 cases, scanned with Aperio ScanScope CS2)</li> <li>Human melanoma (51 cases, scanned with Hamamatsu NanoZoomer XR) (no labels provided)</li> </ul> <p>From each WSI, a trained pathologist selected an area of 2mm² corresponding to approximately 10 high power fields, according to the grading scheme of Elston and Ellis. We cropped this area and provide it as PNG files in this data set due to restrictions in data set size on zenodo. Each file includes the resolution (in dots per inch, DPI) of the original scanned images.</p> <p>The training set contains 9501 mitotic figures (MF) and 11051 hard examples (non-mitotic figures). All annotations are provided in MS COCO JSON format and as SQLITE database (SlideRunner format).</p>
MELA Dataset: A Benchmark for Mediastinal Lesion Analysis (Training Set Part 1)
<p>MELA dataset is a benchmark for developing algorithms on mediastinal lesion analysis. We hope this large-scale dataset could facilitate the research and application of automatic mediastinal lesion detection and diagnosis. </p> <p>MELA dataset contains 1100 CT scans collected from patients with one or more lesions in the mediastinum. The MELA dataset is split into a subset of 770 CT scans for training, a subset of 110 CT scans for validation, and a test set of 220 CT scans for evaluation.</p> <p>Due to the size limit of zenodo.org, we split the MELA training set into 3 parts; this is the Training Set Part 1 of MELA dataset, including 260 CTs. Files include:</p> <ol> <li>Train1.zip: 130 CTs in NII format (nii.gz).</li> <li>Train2.zip: 130 CTs in NII format (nii.gz).</li> </ol>
MELA Dataset: A Benchmark for Mediastinal Lesion Analysis (Training Set Part 2)
<p>MELA dataset is a benchmark for developing algorithms on mediastinal lesion analysis. We hope this large-scale dataset could facilitate the research and application of automatic mediastinal lesion detection and diagnosis. </p> <p>MELA dataset contains 1100 CT scans collected from patients with one or more lesions in the mediastinum. The MELA dataset is split into a subset of 770 CT scans for training, a subset of 110 CT scans for validation, and a test set of 220 CT scans for evaluation.</p> <p>Due to the size limit of zenodo.org, we split the MELA training set into 3 parts; this is the Training Set Part 2 of MELA dataset, including 260 CTs. Files include:</p> <ol> <li>Train3.zip: 130 CTs in NII format (nii.gz).</li> <li>Train4.zip: 130 CTs in NII format (nii.gz).</li> </ol>
MELA Dataset: A Benchmark for Mediastinal Lesion Analysis (Training Set Part 3)
<p>MELA dataset is a benchmark for developing algorithms on mediastinal lesion analysis. We hope this large-scale dataset could facilitate the research and application of automatic mediastinal lesion detection and diagnosis. </p> <p>MELA dataset contains 1100 CT scans collected from patients with one or more lesions in the mediastinum. The MELA dataset is split into a subset of 770 CT scans for training, a subset of 110 CT scans for validation, and a test set of 220 CT scans for evaluation.</p> <p>Due to the size limit of zenodo.org, we split the MELA training set into 3 parts; this is the Training Set Part 3 of MELA dataset, including 250 CTs. Files include:</p> <ol> <li>Train5.zip: 130 CTs in NII format (nii.gz).</li> <li>Train6.zip: 120 CTs in NII format (nii.gz).</li> </ol>
Galaxy Training Data for "Evaluating and ranking a set of pathways based on multiple metrics"
<pre>This dataset provides the inputs needed for the Galaxy Pathway Analysis workflow training tutorial (<a href="https://galaxy-synbiocad.org">https://galaxy-synbiocad.org</a>). This workflow asseses the performance of predicted pathways by computing 4 criteria (target product flux, thermodynamic feasibility, pathway length, and enzyme availability). A score inform the user about the best candidate pathways to produce a compound of interest. The generated output is a collection of scored and ranked heterologous pathways. The content of the dataset is as follows: - A set of pathways provided in the SBML format (Systems Biology Markup Language) to be ranked, modeling heterologous pathways such as those outputted by the RetroSynthesis workflow (<a href="https://galaxy-synbiocad.org">https://galaxy-synbiocad.org</a>). - The GEM (Genome-scale metabolic models) which is a formalized representation of the metabolism of the host organism (the model is E. coli iML1515), provided in the SBML format.</pre>
Fotopersbureau De Boer Training Set on Scene Detection
<p>This dataset was created as part of the projects HisVis and <em><a href="https://noord-hollandsarchief.nl/fotografischgeheugen">Fotografisch Geheugen</a> </em>conducted by the Noord-Hollands Archief and University of Amsterdam in the Netherlands. These projects examined to what extent Computer Vision, and more specifically, scene detection could be applied to a collection of historical press photographs. A computer was trained to automatically recognize specific scenes on historical press photographs, like a ‘protest’, ‘marriage’, ‘shopping street’ or ‘baseball game’. The key aspect of scene recognition is to identify the place in which the objects seat.</p> <p>The specific aim of the enrichments provided by scene detection was to benefit users of the archive and cultural historians studying historical photographs.</p> <p>The training set contains historical press photographs of the collection of Fotopersbureau De Boer (1945-2005). This file includes the training data used for the training of a scene detection model as well as a model to detect whether a picture was taken indoors or outdoors. The model cards, data sheet, and label sheet include more information on the dataset.</p> <p>- examples.tar.gz contains example images for each label. The label sheet contains more information</p> <p>- indoor_out.tar.gz contains the training set for the indoor / outdoor model</p> <p>- scene_detection.tar.gz contains the training set for the scene detection model.</p> <p>- no_description_found.tar.gz contains images that were not linked to any class during annotation.</p> <p>- HisVis2-0.1.beta.tar.gz contains the source code for the software. It stems from this GitHub repo: https://github.com/melvinwevers/HisVis2</p> <p>- models.tar.gz contains the trained models. These should be placed in the repo. The models are not included here because of their size.</p> <p> </p> <p>More information on the data is available in the <a href="https://github.com/melvinwevers/HisVis2/blob/main/docs/data_sheet.md">data sheet</a>:</p> <p> </p> <p><br> </p> <p> </p>
Data from: How many specimens make a sufficient training set for automated three dimensional feature extraction?
<p>Deep learning has emerged as a robust tool for automating feature extraction from 3D images, offering an efficient alternative to labour-intensive and potentially biased manual image segmentation methods. However, there has been limited exploration into the optimal training set sizes, including assessing whether artificial expansion by data augmentation can achieve consistent results in less time and how consistent these benefits are across different types of traits. In this study, we manually segmented 50 planktonic foraminifera specimens from the genus Menardella to determine the minimum number of training images required to produce accurate volumetric and shape data from internal and external structures. The results reveal unsurprisingly that deep learning models improve with a larger number of training images with eight specimens being required to achieve 95% accuracy. Furthermore, data augmentation can enhance network accuracy by up to 8.0%. Notably, predicting both volumetric and shape measurements for the internal structure poses a greater challenge compared to the external structure, due to low contrast differences between different materials and increased geometric complexity. These results provide novel insight into optimal training set sizes for precise image segmentation of diverse traits and highlight the potential of data augmentation for enhancing multivariate feature extraction from 3D images. </p>
Train and Evaluation Code, Road Classification Models and Test set of the paper "Insights into the Effects of Image Overlap and Image Size on Semantic Segmentation Models Trained for Road Surface Area Extraction from Aerial Orthophotography"
<p>This repository contains the Python scripts built for training and evaluation of the implementation, together with the test data and the resulting road segmentation models corresponding to the paper "Insights into the Effects of Image Overlap and Image Size on Semantic Segmentation Models Trained for Road Surface Area Extraction from Aerial Orthophotography". The scripts make use of the Tensorflow with Keras framework and their additional required dependencies.</p> <p>The training and validation set is based on the binary SROADEX dataset (<a href="../records/6482346">https://zenodo.org/records/6482346</a>) that was re-split into tiles that feature the image resolutions (256 x 256, 512 x 512, and 1024 x 1024 pixels) and image overlaps (0% and 12.5%) considered in this study. The data have been generated using scripts developed in Python using Open Source libraries (GDAL/OGR and MapScript) for rasterization of vector cartography that represents the axes of the different types of roads (urban, interurban and rural). This binary road data contains information from 16 full orthoimages (28.5 km * 18.5 km) with spatial resolution of 0.5 m/pixel from the insular and peninsular Spanish territory. Due to the size on disk of approximately 492 gigabytes, this training and validation data is only available upon request from the corresponding author. The test set has been generated from a novel area from Palencia (Spain) and features 18 million pixels labelled with the positive "Road" class. The test sets are provided in the repository for each resolution (with no overlap), so that additional DL models can be evaluated on the same data and compared with the results achieved in this study.</p> <p>The structure of the information shared in this repository is as follows:<br>The scripts have been grouped by tile resolution (256, 512 and 1024). First, the test set and the evaluation script can be found. For each tile resolution, there are two subfolders (corresponding to the "no overlap" and "12.5% overlap"). In each case, the Python scripts for training the models in the three repetitions are shared, and the trained models (H5 format) are shared in compressed form. Finally, for each resolution we also share the testing dataset which consists of two folders.</p> <p>The material is distributed under a CC-BY 4.0 license.</p>
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.