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.
1,273
datasets available to search
ShareScore release 0.9.0
Dataset results
1,273 results for “Deep Learning”
Improving Automatic Melanoma Diagnosis using Deep Learn-ing-based Segmentation of Irregular Networks
<p>Irregular masks dataset created on a subset of the ISIC19 training dataset. All annotations are for melanoma lesions. The filename indicates the ISIC19 image id along with suffix indicating annotator and/or verifier. This dataset was used in the publication "Improving Automatic Melanoma Diagnosis using Deep Learn-ing-based Segmentation of Irregular Networks" to be submitted to the Cancers Journal.</p> <p>Please cite the corresponding article (to be published) if data is used in your work.</p> <p>The references for the ISIC19 dataset that this is built on is given below.</p> <blockquote> <p>BCN_20000 Dataset: (c) Department of Dermatology, Hospital Clínic de Barcelona</p> <p>HAM10000 Dataset: (c) by ViDIR Group, Department of Dermatology, Medical University of Vienna; <a href="https://doi.org/10.1038/sdata.2018.161">https://doi.org/10.1038/sdata.2018.161</a></p> <p>MSK Dataset: (c) Anonymous; <a href="https://arxiv.org/abs/1710.05006">https://arxiv.org/abs/1710.05006</a>; <a href="https://arxiv.org/abs/1902.03368">https://arxiv.org/abs/1902.03368</a></p> </blockquote>
Detecting coarse beach sediment using remotely sensed imagery at the FRF, Duck, NC, USA: Labeled images, deep learning model, testing data, and predictions.
<p>This data record contains 5 zip files all used to build and use a semantic segmentation model to operate on beach imagery taken at the Field Research Facility (FRF) in Duck, North Carolina, USA. All data is from 2015-2021</p> <p>The `training_data.zip` contains all data used to train the ML model. All images come from the north facing (c1) camera. This zip file includes: a list of classes used to label the imagery, and folders of 107 images, 107 sparse annotations (doodles), 107 labels, and 107 overlays. All labeling was done with the open-source labeling tool ‘Doodler (Buscombe et al., 2021).</p> <p>The `model.zip` file contains the ML model, and associated metadata. This includes: a JSON model configuration file, a figure showing model training statistics, an `.npz` file of model training output, a list of training and validation files, the model as an h5 file and in the Tensorflow ‘saved model’ format. All modeling was done with Segmentation Gym (Buscombe & Goldstein 2022).</p> <p>The `test_data_c6.zip` file contains all data from the south facing (c6) camera to test the ML model. This includes: a list of classes used to label the imagery, and folders of 10 images, 10 sparse annotations (doodles), 10 labels, and 10 overlays. All labeling was done with the open-source labeling tool ‘Doodler (Buscombe et al., 2021). Testing the model with this data was done with codes in: https://github.com/ebgoldstein/FRF_GrainSize</p> <p>The `test_data_c1.zip` file contains all data from the north facing (c1) camera to test the ML model. This includes: a list of classes used to label the imagery, and folders of 10 images, 10 sparse annotations (doodles), 10 labels, and 10 overlays. All labeling was done with an open-source labeling tool ‘Doodler (Buscombe et al., 2021). Testing the model with this data was done with codes in: https://github.com/ebgoldstein/FRF_GrainSize</p> <p>The `predictions.zip` file contains 4418 images from the north facing (c1) camera that were run through the trained segmentation model as well as the resulting output (presented as side-by-side image and overlays). These images were created using codes in Segmentation Gym (Buscombe & Goldstein 2022).</p>
Data archive for Exploiting radar polarimetry for nowcasting thunderstorm hazards using deep learning
<p>This dataset contains the machine learning training data files, pretrained model weights and results for the paper Exploiting radar polarimetry for nowcasting thunderstorm hazards using deep learning, submitted to Natural Hazards and Earth System Sciences, 2023.</p> <p>The radar dataset can be found at the following Zenodo repository: <a href="https://doi.org/10.5281/zenodo.6325370">https://doi.org/10.5281/zenodo.6325370</a></p> <p>For instructions for using the data, please see the GitHub code repository at <a href="http://github.com/meteoswiss/c4dl-polar">https://github.com/meteoswiss/c4dl-polar</a>. Download all the files here and extract the contents to the following subdirectories in the ML code directory:</p> <ul> <li>Training data (patches_quality-index_2020.zip or patches_*_2020.nc) -> data/2020/</li> <li>Results: (results.zip) -> runs/run*/results/</li> <li>Pretrained models (models_run*) -> runs/run*/</li> </ul>
Supporting data for "CoVEffect: Interactive System for Mining the Effects of SARS-CoV-2 Mutations and Variants Based on Deep Learning"
<p>This repository contains the datasets created and extracted for the paper:</p> <p>Giuseppe Serna García, Ruba Al Khalaf, Francesco Invernici, Stefano Ceri, and Anna Bernasconi. 2022.<br> "<strong>CoVEffect</strong>: Interactive System for Mining the <strong>Effects of SARS-CoV-2 Mutations and Variants</strong> Based on Deep Learning". (Available online at http://gmql.eu/coveffect)</p> <p>--------------------------------------------------------------------------------<br> LIST OF FILES WITH DESCRIPTION:<br> --------------------------------------------------------------------------------</p> <p>AdditionalFile1-effects-taxonomy:<br> Descriptions of legal values for the 'Effect' field, based on a categorized taxonomy.</p> <p>AdditionalFile2-levels-taxonomy:<br> Descriptions of legal values for the 'Level' field.</p> <p>AdditionalFile3-training_dataset_target:<br> List of target tuples (manually annotated) of 221 abstracts considered for training the model. For each abstract, target tuples follow the schema ID, DOI, title, entity, effect, level, type (mutation or variant), tuples_count (>1 when an effect/level is shared by multiple entities, #abstracts containing the same effect described in the tuple).</p> <p>AdditionalFile4-validation_dataset_target:<br> List of target tuples (manually annotated) of 50 abstracts considered for validating the prepared prediction model.<br> For each abstract, target tuples follow the schema defined for AdditionalFile3.</p> <p>AdditionalFile5-validation_dataset_highlighted:<br> Textual abstracts of the 50 manuscripts considered for validation; the text used to support the manual target annotations has been highlighted in yellow.</p> <p>AdditionalFile6-validation_dataset_prediction:<br> List of predicted annotations of 50 abstracts considered for validating the prepared prediction model. The file is split in 4 TSV, respectively for entity (a), effect (b), level (c), and whole tuple predictions (d).</p> <p>AdditionalFile7-keywords_query_list:<br> Keyword-based search run on the CORD-19 dataset to extract a relevant subset of abstracts regarding the scope of interest of CoVEffect. The Boolean logic used to combine keywords is explained in the section 'Annotations of the biology-related CORD-19 cluster'.</p> <p>AdditionalFile8-CORD-19_batch_dataset_metadata:<br> Metadata of the 7,230 papers extracted by the keyword-based query in AdditionalFile7.<br> These abstracts have been annotated by the prediction framework.</p> <p>AdditionalFile9-CORD-19_batch_dataset_prediction:<br> List of predicted annotations of 7,230 abstracts extracted from the biology-related cluster of CORD-19.</p> <p>AdditionalFile10-test_dataset_target:<br> List of target tuples (manually annotated) of 100 abstracts randomly selected from the 7,230 extracted as in AdditionalFile8.<br> For each abstract, target tuples follow the schema defined for AdditionalFile3.</p> <p>AdditionalFile11-test_dataset_prediction:<br> List of predicted annotations of 100 abstracts considered for testing the prediction model on a subset of the CORD-19 biology-related cluster. As AdditionalFile6, it is split in 4 TSV, respectively for entity (a), effect (b), level (c), and whole tuple predictions (d).</p>
Multicenter Validated Detection of Focal Cortical Dysplasia using Deep Learning
<p>Lesional and non-lesional patches derived from 148 FCD patients is available as a HDF5 dataset (v1.0.0; doi: 10.5061/dryad.h70rxwdgm or 10.5281/zenodo.3239446). To create this dataset, for each of the 148 FCD patients, we sampled at most 1,000 (*_N1000.h5) or 1,500 (*_N1500.h5) cortical patches (or # voxels in the lesion, whichever is lower) of size 16×16×16 within the lesion.The same number of cortical patches were sampled randomly outside the lesion. The resulting lesional and non-lesional patches were concatenated, shuffled (to add another layer of randomization), and saved along with their binary labels (compressed HDF5 dataset).</p> <p>For axis=1, index 0 is T1 and 1 is FLAIR.</p>
Inverse design of metal-organic frameworks for direct air capture of CO2 via deep reinforcement learning
<p>The combination of several interesting characteristics makes metal-organic frameworks (MOFs) a highly sought-after class of nanomaterials for a broad range of applications like gas storage and separation, catalysis, drug delivery, and so on. However, the ever-expanding and nearly infinite chemical space of MOFs makes it extremely challenging to identify the most optimal materials for a given application. In this work, we present a novel approach using deep reinforcement learning for the inverse design of MOFs, our motivation being designing promising materials for the important environmental application of direct air capture of CO2 (DAC). We demonstrate that the reinforcement learning framework can successfully design MOFs with critical characteristics important for DAC. Our top-performing structures populate two separate subspaces of the MOF chemical space: the subspace with high CO2 heat of adsorption and the subspace with preferential adsorption of CO2 from humid air, with few structures having both characteristics. Our model can thus serve as an essential tool for the rational design and discovery of materials for different target properties and applications.</p>
A deep learning-based dataset of WFA-positive perineuronal nets and parvalbumin neurons localizations in the adult mouse brain
<p><strong>Quality-controlled predictions of deep learning models for cell counting</strong></p> <p>This dataset contains high-resolution images for the visualization of perineuronal nets (PNNs) and parvalbumin-expressing (PV) cells analyzed in the paper:</p> <p><em>A Comprehensive Atlas of Perineuronal Net Distribution and Colocalization with Parvalbumin in the Adult Mouse Brain.</em></p> <p>The dataset integrates the raw data published on a <a href="https://zenodo.org/record/7419282">previous upload</a> on Zenodo.</p> <p>Cell locations were obtained using two deep-learning models for cell counting (publicly available on <a href="http://github.com/ciampluca/counting_perineuronal_nets">GitHub</a>, details in the paper by <a href="https://www.sciencedirect.com/science/article/pii/S1361841522001475">Ciampi et al., 2022</a>). The output of the deep-learning pipeline was filtered based on the <em>score</em> assigned to each cell prediction, by removing all the PNNs with a score lower than 0.4 and all the PV cells with a score lower than 0.55. Cases of artefactual cell detection were finally removed manually by visual inspection of the images. </p> <p><strong>Content</strong></p> <p>The dataset contains microscopy images of coronal brain slices from 7 adult mice. The objects highlighted in these images represent the final set of PNNs/PV cells that were used in all the analysis of the paper.</p> <p><strong>Folder Structure and file naming conventions</strong></p> <p>There are separate folders for each mouse. Each folder is named with the ID of that mouse. Within each folder, images are assigned a code specifying the channel (C1 for PNNs, C2 for PV cells).</p> <p> </p>
FORMS: Forest Multiple Source height, wood volume, and biomass maps in France at 10 to 30 m resolution based on Sentinel-1, Sentinel-2, and GEDI data with a deep learning approach.
<p>The products can be vizualized at <a href="https://martinschwartz0.users.earthengine.app/view/forms-height-biomass-volume-viewer">https://martinschwartz0.users.earthengine.app/view/forms-height-biomass-volume-viewer</a></p> <p>- FORMS-H: Canopy height map of France at 10 m resolution. The units are in centimeter (10^-2 m).</p> <p>- FORMS-B: Above-ground biomass density map of France at 30 m resolution. The units are in Mg ha-1</p> <p>- FORMS-V: Wood volume density map of France at 30 m resolution. The units are in m3 ha-1</p> <p>Please refer to the paper <a href="https://doi.org/10.5194/essd-15-4927-2023">https://doi.org/10.5194/essd-15-4927-2023</a> for further details.</p>
Phononic crystals dataset for supervised training of surrogate deep learning model
<p>The dataset contains shapes of unit cells of phononic crystals (inputs) in the form of images and corresponding dispersion diagrams (outputs). The dataset is used for deep learning (DL) model training.<br> Outputs are in the form of .mat files which contain vectors of reduced wavevector and corresponding frequencies, and also displacements u, v, w which can be used for polarization calculation.</p> <p>The dataset contains 11000 cases.</p> <p>Note: Ignore names "labels" as these are actually inputs to the DL model, not labels.</p>
A Novel Approach to Heart Failure Prediction and Classification through Advanced Deep Learning Model
<p>A Novel Approach to Heart Failure Prediction and Classification through Advanced Deep Learning Model</p>
Modeling islet enhancers using deep learning identifies candidate causal variants at loci associated with T2D and glycemic traits
<p>Genetic association studies have identified hundreds of independent genetic signals associated with type 2 diabetes (T2D) and related traits. Despite these successes, the identification of specific causal variants underlying a genetic association signal remains challenging. In this study, we describe a deep learning method to analyze the impact of sequence variants on enhancers. Focusing on pancreatic islets, a relevant T2D tissue, we show that our model learns islet-specific transcription factor (TF) regulatory patterns and can be used to prioritize candidate causal variants. At 101 genetic signals associated with T2D and related glycemic traits where multiple variants occur in linkage disequilibrium, our method nominates a single causal variant for each association signal, including three variants previously shown to alter reporter activity in islet-relevant cell types. For another signal associated with blood glucose levels, we biochemically test all candidate causal variants from statistical fine-mapping using a pancreatic islet beta cell line and show biochemical evidence of allelic effects on TF binding for the model-prioritized variant. To aid in future research, we publicly distribute our model and islet enhancer perturbation scores across ~67 million variants. We anticipate that deep learning methods like the one presented in this study will enhance the prioritization of candidate causal variants for functional studies.</p>
LocScale-EMmerNet deep learning models for contrast optimisation of cryo-EM maps
<p>EMmerNet deep learning models for local optimisation of cryo-EM map contrast using <a href="https://gitlab.tudelft.nl/aj-lab/locscale">LocScale</a>.</p>
UBGG-3m: Fine-grained urban blue-green-gray landscape dataset for 36 Chinese cities based on deep learning network
<p>The UBGG dataset provides easily access and leverage to researchers and analysts, which is stored in the following Zenodo repository (<a href="https://doi.org/10.5281/zenodo.8053333">https://doi.org/10.5281/zenodo.8352777</a>). The UBGG dataset consists of two main components:</p> <ul> <li><strong>UBGG-3m: the fine-grained UBGG map product of 36 metropolises in China.</strong> The UBGG-3m dataset captures the intricate urban landscape features with remarkable precision, providing a detailed representation at an impressive 3-meter resolution. Fig. 1 in User Guides shows the classification results for 36 Chinese metropolises. Researchers can delve into the nuances of the UBGG continuum, gaining invaluable insights into the interplay between the blue, green, and gray elements of urban environments in each metropolis.</li> </ul> <ul> <li><strong>UBGGset:</strong> <strong>the large-volume sample dataset to support the UBGG deep learning research.</strong> Complementing the UBGG-3m dataset, UBGGset serves as a large-volume sample dataset specifically tailored to support and foster UBGG research endeavors (Fig. 2). The UBGGset consists of 14,627 sample images (without data augmentation), with dimensions of 256 pixels in length and width, covering an urban area of approximately 2,272 km<sup>2</sup>. The UBGGset was constructed with co-registered pairs of 3 m Planet images and fine-annotated urban landscapes labeled on 1 m Google Earth image. This dataset encompasses 15 typical cities, offering researchers a rich and diverse resource to drive exploration, analysis, and innovation in the field of urban landscape studies.</li> </ul> <p> </p> <p><strong>Citation format for paper and dataset:</strong></p> <p>[1] Zhiyu Xu, Shuqing Zhao. Fine-grained urban blue-green-gray landscape dataset for 36 Chinese cities based on deep learning network. <em>Sci Data</em> 11, 266 (2024). https://doi.org/10.1038/s41597-023-02844-2</p> <p>[2] Zhiyu Xu, Shuqing Zhao, Fine-grained urban landscape mapping reveals broad-scale homogeneity in urban environments,<br>Science Bulletin, (2024). https://doi.org/10.1016/j.scib.2024.03.060</p> <p>[3] Zhiyu Xu, Shuqing Zhao. UBGG-3m: Fine-grained urban blue-green-gray landscape dataset for 36 Chinese cities based on deep learning network (v1.0) [Data set]. (2023). Zenodo. https://doi.org/10.5281/zenodo.8352777</p>
Dataset of Real Faults in Deep Learning Systems
<p>This is a replication package for the "Taxonomy of Real Faults in Deep Learning Systems" paper.</p> <p>The dataset contains information on all the issues and real faults gathered in the course of the study.</p> <p>The dataset consists of three main folders: Manual_Labelling, Interviews and Survey.</p> <p>Manual_Labelling</p> <p>In this folder we have placed all the files associated with our manual labelling process. It is divided in 3 subfolders: SO_init, GitHub_init and Analysed_Artifacts.</p> <p>GitHub_init</p> <p>The GitHub mining process is explained in detail in Section 3.1.1 of the paper. The list of initially mined GitHub projects for each framework is presented in the file {frameworkname}_init.csv. Each line in the file provides project name, the link to the project and its general information such as number of commits, issues, stars and etc. As part of our procedure, we removed projects that do not represent real software systems. For such projects we provide an explanation on why it should be excluded in the column "Comment". The files {frameworkname}_after_init_removal.csv present the list of remaining projects after the exclusion of such systems. The files {frameworkname}_top_100.csv list the top 100 projects we selected for our final analysis.</p> <p>As explained in Section 3.1.1, to identify relevant commits and issues we used a vocabulary of related terms. The complete list of 11,968 stemmed words is presented in the file vocabulary_init.csv. The final list of 105 relevant words obtained by the exclusion of the words that appear less than 10 times and the further manual analysis is provided in the file vocabulary_final.csv.</p> <p>SO_init</p> <p>The extraction procedure from StackOverflow is explained in detail in Section 3.1.2 of the paper. In the file query.rtf we provide the code of the query we have used to extract discussions related to each of the analysed frameworks (Keras, Torch, Tensorflow) from StackOverflow. As a result, we got a csv file that lists discussions for each of the frameworks (files keras.csv, torch.csv and tensorflow.csv).</p> <p>Analysed_Artifacts</p> <p>Our manual analysis was conducted in 6 rounds (Table 1 in the paper). We provide information about each round in a separate csv file round_{roundnumber}.csv. In each file we provide the following information: (1) <strong>Entity Type</strong> - whether it is a StackOverflow or GitHub artifact and which framework it corresponds to; (2) <strong>Link</strong> to the artifact; (3) <strong>Number of Evaluators</strong> - how many evaluators were assigned to this artifact; (4) <strong>Conflict</strong> - whether there was a conflict between evaluators when assigning tag to this artifact, has a value 0 for no and 1 for yes; (5) <strong>Evaluator1..4</strong> and <strong>Tag1..4</strong> - the ID of each evaluator and the tag provided by each of the assigned evaluators; (6) <strong>Final Tag</strong> - final tag assigned to the artifact; (7) <strong>Taxonomy Tag</strong> - the tag in the final taxonomy;</p> <p>The detailed information on the statistics of each round can be found in file stats.xlsx.</p> <p>Interviews</p> <p>The background information about our interview participants is presented in the file interview_participant_info.xlsx. The interview guide we used to conduct the semi-structured interviews is in the file interview_guide.docx. In the subfolder "Transcriptions" we provide transcribed versions of all 20 interviews.</p> <p>We provide the details of open coding process for the interviews in the file interview_open_coding.xlsx. Each row corresponds to a part of the interview text to which at least one of the evaluators has assigned a tag. Therefore, each row contains the following information: (1) <strong>Interview Num</strong> - the interview number; (2) <strong>Evaluator 1</strong> - tag provided by the first evaluator; (3) <strong>Evaluator 2</strong> - tag provided by the second evaluator; (4) <strong>Moderator Tag</strong> - tag assigned by the moderator; (5) <strong>Tag</strong> - the final taxonomy tag; (6) <strong>Status</strong> - whether this tag has been added to the final taxonomy or not ("A" for yes and "R" for no).</p> <p>In the file Interview_Tags.xlsx we provide aggregated information about the tags obtained from interviews. The column "Number" shows the number of times the tag was extracted from the interviews. Similarly to the previous file, the column "Status" shows whether the tag became part of the final taxonomy or not.</p> <p>Survey</p> <p>We provide the survey form we have used for our validation study in the file survey_form.pdf. The file with the information about participants and their answers to the survey questions are in the file participant_info_and_responses.xlsx. The percentages reported in Table 2 in the paper are calculated in this file (last rows with a bold font).</p>
LOFAR dataset for deep learning assisted data Inspection for radio astronomy
<p>This dataset is used for the training of the magnitude and phase-based VAE in the paper entitled "Deep learning assisted data inspection for radio astronomy".</p> <p>For uploading purposes the dataset has been separated into 4 different .zip files. In order to use this dataset each of the zip files should be extracted into a single directory so that the training can be performed on all files at the same time. <br> <br> More information can be found on <a href="https://github.com/mesarcik/DL4DI">the project github repository</a>. </p>
Counted Biopores dataset used in 'RootPainter: Deep Learning Segmentation of Biological Images with Corrective Annotation'
<p>Counted biopores dataset used in the article: 'RootPainter: Deep Learning Segmentation of Biological Images with Corrective Annotation'</p> <p>Originally collected as part of a field trial at the University of Bonn in 2012, described in the following paper:</p> <p>Eusun Han, Timo Kautz, Ute Perkons, Marcel Lüsebrink, Ralf Pude, and Ulrich Köpke.Quantification of soil biopore density after perennial fodder cropping.Plant and Soil, 394(1-2):73–85, sep 2015. ISSN 15735036. doi:10.1007/s11104- 015- 2488- 3</p>
Counted Nodules dataset used in 'RootPainter: Deep Learning Segmentation of Biological Images with Corrective Annotation'
<p>Counted Nodules dataset used in the article: 'RootPainter: Deep Learning Segmentation of Biological Images with Corrective Annotation'</p>
Dataset: Whole blood count, used in: "AIDeveloper: deep learning image classification in life science and beyond"
<p>Real-time deformability cytometry (RT-DC) data of whole blood measurements.<br> Data was used to train and validate a neural net to perform a blood count based on brightfield images of RT-DC.</p> <p>01_Model: Contains the final model as well as an AIDeveloper meta-file that allows to reproduce the training procedure. The metafile preciesely defines which dataset was used for training and which for validation as well as all parameters that were set in AIDeveloper.</p> <p>The following folders contain data that was used for training (and validation):</p> <ul> <li>Cambr</li> <li>KIK</li> <li>20190306_DextranBlood_AI_DataSet</li> <li>Gs_Blood_Train</li> </ul> <p>Testing data is stored on figshare:<br> https://figshare.com/articles/Krater_et_al_2020_Data_zip/9902636</p>
cigKast: A data of 3D synthetic seismic volumes with labeled paleokarsts for deep-learning-based paleokarst interpretation
<p>cigKarst is a dataset created by the <a href="http://cig.ustc.edu.cn/">Computational Interpretation Group (CIG)</a> for the deep-learning-based peleokarst interpretation in 3D seismic images, <a href="http://cig.ustc.edu.cn/xinming/list.htm" target="_blank" rel="noopener">Xinming Wu</a> is the main contributor to the dataset.</p> <p>This dataset contains 120 pairs of synthetic 3D seismic images and the corresponding label images with the ground truth of the paleokarst systems simulated in the seismic images. More detail of building this dataset is discussed in the paper published at the journal of JGR Solid Earth:</p> <p><strong>Wu, X.</strong>, S. Yan, J. Qi, and H. Zeng, 2020, Deep learning for characterizing paleokarst collapse features in 3D seismic images. <strong>JGR, Solid Earth</strong>, Vol. 125(9), 1-23, e2020JB019685. <a href="http://cig.ustc.edu.cn/_upload/tpl/05/cd/1485/template1485/papers/wu2020karst.pdf">[PDF]</a>. doi: 10.1029/2020JB019685</p> <p>Below are some brief description of the dataset:</p> <p>1) The "seismic.zip" contains 120 3D seismic images, each image is with the dimension of 256X256X256;</p> <p> 2) The "karst.zip" contains 120 3D label images of the karsts. Each label image is with the same dimension of 256X256X256. The values in a label image are set with ones in the karst areas while zeros elsewhere, which is why the compressed label images in the karst.zip is much smaller than the seismic images compressed in the seismic.zip</p>
Shared Acoustic Codes Underlie Emotional Communication in Music and Speech - Evidence from Deep Transfer Learning (Datasets)
<p>This repository contains the datasets used in the article "Shared Acoustic Codes Underlie Emotional Communication in Music and Speech - Evidence from Deep Transfer Learning" (Coutinho & Schuller, 2017). </p> <p>In that article four different data sets were used: SEMAINE, RECOLA, ME14 and MP (acronyms and datasets described below). The SEMAINE (speech) and ME14 (music) corpora were used for the unsupervised training of the Denoising Auto-encoders (domain adaptation stage) - only the audio features extracted from the audio files in these corpora were used and are provided in this repository. The RECOLA (speech) and MP (music) corpora were used for the supervised training phase - both the audio features extracted from the audio files and the Arousal and Valence annotations were used. In this repository, we provide the audio features extracted from the audio files for both corpora, and Arousal and Valence annotations for some of the music datasets (those that the author of this repository is the data curator).</p> <p>Below, you can find description of the various corpora, the details about the data stored in this repository and information on how to obtain the rest of the data used by Coutinho and Schuller (2017).</p> <p><strong>SEMAINE (speech)</strong></p> <p>The SEMAINE corpus (McKeown, Valstar, Cowie, Pantic & Schroder, 2012) was developed specifically to address the task of achieving emotion-rich interactions, and it is adequate for this task as it comprises a wide range of emotional speech. It includes video and speech recordings of spontaneous interactions between human and emotionally stereotyped `characters'. Coutinho & Schuller (2017) used a subset of this database (called <em>Solid-SAL</em>). The <em>Solid-SAL</em> dataset is freely available for scientific research purposes (see http://semaine-db.eu). This repository includes the audio features used in Coutinho & Schuller (2017) (under features/SEMAINE).</p> <p><strong>RECOLA (speech)</strong></p> <p>The RECOLA database (Ringeval, Sonderegger, Sauer & Lalanne, 2013) consists of multimodal recordings (audio, video, and peripheral physiological activity) of spontaneous dyadic interactions between French adults. Coutinho & Schuller (2017) used the RECOLA-Audio module which consists of the audio recordings of each participant in the dyadic phase of the task. In particular, they used the non-segmented high-quality audio signals (WAV format, 44.1kHz, 16bits), obtained through unidirectional headset microphones, of the first five minutes of each interaction. Annotations consist of time-continuous ratings of the level of Arousal and Valence dimensions of emotion perceived by each rater while seeing and listening the audio-visual recordings of each participant task. The publicly available annotated dataset includes only part of the data which amounts to a total number of 23 instances. The time frame length used by Coutinho & Schuller (2017) is 1s (the original annotations were downsampled). This repository includes the audio features used in Coutinho & Schuller (2017) (under features/RECOLA). To obtain the annotations you should contact the author of the original study (see https://diuf.unifr.ch/diva/recola/download.html for further details).</p> <p><strong>ME14 (music)</strong></p> <p>The MediaEval ``Emotion in Music'' task is dedicated to the estimation of Arousal and Valence scores continuously in time and value for song excerpts from the Free Music Archive. Coutinho and Schuller (2017) used the whole corpus (development and test sets for the 2014 challenge) which includes 1,744 songs belonging to 11 musical styles -- Soul, Blues, Electronic, Rock, Classical, Hip-Hop, International, Folk, Jazz, Country, and Pop (maximum of five songs per artist). This repository includes the audio features used in Coutinho & Schuller (2017) (under features/ME14). The full dataset (including annotations) can be obtained from http://www.multimediaeval.org/mediaeval2014/emotion2014/.</p> <p><strong>MP (music)</strong></p> <p>This is a corpus compiled specifically for this work described in Coutinho & Schuller (2017) using data collected in four previous studies. It consists of emotionally diverse full music pieces from a variety of musical styles (Classical and contemporary Western Art, Baroque, Bossa Nova, Rock, Pop, Heavy Metal, and Film Music). Annotations were obtained in controlled laboratory experiments whereby the emotional character of each piece was evaluated time-continuously in terms of levels of Arousal and Valence perceived by listeners (ranging between 35 to 52 in the four studies). In what follows, some details about the various studies are described.</p> <ul> <li>MP<sub>DB1</sub>: This subset of the MP corpus consists of the data reported by Korhonen (2004), and gently made available by the author. This dataset includes six full (or long excerpts) music pieces ranging from 151s to 315s in length (only classical music). Each piece was annotated by 35 participants (14 females). The time series correspondents to each music piece were collected at 1Hz. The golden standard for each piece was computed by averaging the individual time series across all raters. This repository includes the audio features used in Coutinho & Schuller (2017) (under features/MP/DB1). To obtain the labels please contact the author of the original study.</li> <li>MP<sub>DB2</sub>: The dataset by Coutinho & Cangelosi (2011) includes 9 full pieces (43s to 240s long) of classical music (romantic repertoire) annotated by 39 subjects (19 females). Values were recorded every time the mouse was moved with a precision of 1 ms. The resultant timeseries were then resampled (moving average) to a synchronous rate of 1 Hz. The golden standard for each piece was computed by averaging the individual time series across all raters. This repository includes the audio features (under features/MP/DB2) and labels (under annotations/MP/DB2) used in Coutinho & Schuller (2017).</li> <li>MP<sub>DB3</sub>: This dataset was collected by Coutinho & Dibben (2012) and it consists of 8 pieces of film music (84s to 130s long) taken from the late 20th century Hollywood film repertoire. Emotion ratings were given by 52 participants (26 females). The annotation procedure, data processing, and golden standard calculations were identical to MP<sub>DB2</sub>. This repository includes the audio features (under features/MP/DB3) and labels (under annotations/MP/DB3) used in Coutinho & Schuller (2017).</li> <li>MP<sub>DB4</sub>: This dataset was collected by Grewe, Nagel, Kopiez and Altenmüller (2007), and gently made available by the authors. It includes seven music pieces (127s to 502s in length) of heterogeneous styles (e.g., Rock, Pop, Heavy Metal, Classical). Each music piece was annotated by 38 participants (29 females) using an identical methodology to MP<sub>DB2</sub> and MP<sub>DB3</sub>. Data processing and golden standard calculations were also identical. This repository includes the audio features (under features/MP/DB4) used in Coutinho & Schuller (2017). To obtain the labels contact the authors of the original study</li> </ul> <p> </p> <p><strong>Bibliography</strong></p> <p>Coutinho, E., & Cangelosi, A. (2011). Musical emotions: predicting second-by-second subjective feelings of emotion from low-level psychoacoustic features and physiological measurements. <em>Emotion</em>, <em>11</em>(4), 921.</p> <p>Coutinho, E., & Dibben, N. (2013). Psychoacoustic cues to emotion in speech prosody and music. <em>Cognition & Emotion</em>, <em>27</em>(4), 658-684.</p> <p>Coutinho E, Schuller B (2017) Shared acoustic codes underlie emotional communication in music and speech—Evidence from deep transfer learning. PLoS ONE 12(6): e0179289. https://doi. org/10.1371/journal.pone.0179289.</p> <p>Grewe, O., Nagel, F., Kopiez, R., Altenmüller, E. (2007). Emotions over time: synchronicity and development of subjective, physiological, and facial affective reactions to music. <em>Emotion, 7</em>(4), pp. 774-788. DOI: 10.1037/1528-3542.7.4.774.</p> <p>Korhonen, M. (2004). Modeling Continuous Emotional Appraisals of Music Using System Identification. Available from: http://hdl.handle.net/10012/879.</p> <p>McKeown, G., Valstar, M., Cowie, R., Pantic, M., Schroder, M. (2012). The SEMAINE Database: Annotated Multimodal Records of Emotionally Colored Conversations between a Person and a Limited Agent. <em>IEEE Transactions on Affective Computing</em>, 3, pp. 5-17. DOI: http://doi.ieeecomputersociety.org/10.1109/T-AFFC.2011.20.</p> <p>Ringeval, F., Sonderegger, A., Sauer, J. & Lalanne, D. (2013). Introducing the RECOLA Multimodal Corpus of Remote Collaborative and Affective Interactions. In <em>Proceedings of the 2nd International Workshop on Emotion Representation, Analysis and Synthesis in Continuous Time and Space (EmoSPACE 2013)</em>, Shanghai, China. IEEE</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.