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.
129
datasets available to search
ShareScore release 0.9.0
Dataset results
129 results for “Deep neural network”
Supplementary Material: Exposing Previously Undetectable Faults in Deep Neural Networks
<p>Supplementary material for ISSTA 2021 submission #58. The supplementary material is under 600MB and uploaded before the submission deadline, but uploading to the submission website did not work. Editing this supplementary material is not possible after upload.</p>
multi class dataset_Dipper Throated Optimization with Deep Convolutional Neural Network-based Crop Classification on Remote Sensing Image Analysis
Open the record for dataset details and reuse information.
Emergence of Emotion Selectivity in Deep Neural Networks Trained to Recognize Visual Objects
<h1>Datasets and analysis code of the following publication:</h1> <p>Peng Liu, Ke Bo, Mingzhou Ding and Ruogu Fang (2024). Emergence of Emotion Selectivity in Deep Neural Networks Trained to Recognize Visual Objects. <em>PLOS Computational Biology. </em>DOI: 10.1371/journal.pcbi.1011943</p> <p>For any questions please contact the first author at mail pliu1 [at] ufl [dot] edu</p> <h2><strong>Contents:</strong></h2> <p><strong> Code_DataAnalysis</strong><br> - Extracted Selectivity</p> <p> -- IAPS and NAPS datasets</p> <p> -- Neurons In Alexnet and VGG networks</p> <p> --Networks are pre-trained on ImageNet and randomly initialized</p> <p> - Extracted Overlapped Selectivity across IAPS and NAPS.</p> <p> - Extracted tuning performance changes from two datasets and the VGG network</p> <p> -Code to replicate the key results including </p> <p> --Tuning quality </p> <p> -- Number of overlapped neurons</p> <p> -- Enhance neuron activity</p> <p> -- Lesion neurons</p> <p> <strong>TrainedNetworks</strong></p> <p> --Pre-trained VGG network on ImageNet </p> <p> --Pre-trained Alexnet network on ImageNet</p> <p>After pre-training these networks on ImageNet, we fixed their weights and trained them to classify pleasant, neutral, and unpleasant images into three emotion categories using both IAPS and NAPS datasets.<br> </p> <p><strong>Image datasets</strong></p> <p>Access image datasets by request from https://csea.phhp.ufl.edu/media/iapsmessage.html for IAPS and https://lobi.nencki.edu.pl/research/8/ for NAPS.</p>
Data used in "Biologically informed deep neural network for prostate cancer discovery" publication
<p>Data used in the publication titled "<strong>Biologically informed deep neural network for prostate cancer discovery </strong>" </p> <p>Elmarakeby, Haitham A., et al. "Biologically informed deep neural network for prostate cancer discovery." <em>Nature</em> 598.7880 (2021): 348-352.</p> <p>These datasets were derived from the following public domain resources:</p> <ol> <li>Armenia J, Wankowicz SAM, Liu D, Gao J, Kundra R, Reznik E, et al. The long tail of oncogenic drivers in prostate cancer. Nat Genet. 2018;50: 645–651. DOI: <a href="https://doi.org/10.1038/s41588-018-0078-z">10.1038/s41588-018-0078-z</a></li> <li>Fraser M, Sabelnykova VY, Yamaguchi TN, Heisler LE, Livingstone J, Huang V, et al. Genomic hallmarks of localized, non-indolent prostate cancer. Nature. 2017;541: 359–364. https://doi.org/10.1038/nature20788</li> <li>Robinson DR, Wu Y-M, Lonigro RJ, Vats P, Cobain E, Everett J, et al. Integrative clinical genomics of metastatic cancer. Nature. 2017;548: 297–303. https://doi.org/10.1038/nature23306</li> <li>Fabregat A, Jupe S, Matthews L, Sidiropoulos K, Gillespie M, Garapati P, et al. The Reactome Pathway Knowledgebase. Nucleic Acids Res. 2018;46: D649–D655. DOI: <a href="https://doi.org/10.1093/nar/gkv1351">10.1093/nar/gkv1351</a></li> </ol> <p> </p>
Fortran/Python Interface in ARP-GEM1: Online Test of Neural Network Deep Convection
<p>Manuscript under review in AIES. Supporting Code and Dataset. </p> <p><strong>Abstract.</strong></p> <p>In this study, we present the integration of a neural network-based parameterization into the global atmospheric model ARP-GEM1, leveraging the Python interface of the OASIS coupler. This approach facilitates the exchange of fields between the Fortran-based ARP-GEM1 model and a Python component responsible for neural network inference. As a proof-of-concept experiment, we trained a neural network to emulate the deep convection parameterization of ARP-GEM1. Using the flexible Fortran/Python interface, we have successfully replaced ARP-GEM1's deep convection scheme with a neural network emulator. To assess the performance of the neural network deep convection scheme, we have run a 5-years ARP-GEM1 simulation where the neural network replaced ARP-GEM1's deep convection parameterization. The evaluation of averaged fields showed good agreement with output from an ARP-GEM1 simulation using the physics-based deep convection scheme. The Python component was deployed on a separate partition from the general circulation model, using GPUs to increase inference speed of the neural network.</p>
Dataset for "Face Detection in Untrained Deep Neural Networks"
<p><strong>Dataset for</strong></p> <p><strong>"Face Detection in Untrained Deep Neural Networks"</strong></p> <p>Seungdae Baek, Min Song, Jaeson Jang, Gwangsu Kim, and Se-Bum Paik*</p> <p>*Contact: <a href="mailto:sbpaik@kaist.ac.kr">sbpaik@kaist.ac.kr</a></p> <p> </p> <p>To run demo codes for "Face Detection in Untrained Deep Neural Networks", please download files below.</p> <p> </p> <p><strong>1. Stimulus.zip</strong></p> <p>- <strong>IMG_cntr_210521.mat</strong> : A low-level feature-controlled stimulus set was used to find units that responded selectively to face images (Stigliani, 2015). Specifically, 260 images were prepared for each class (face, hand, horn, flower, chair, and scrambled face).</p> <p>- <strong>IMG_var_pos/size/rot_210521.mat</strong> : To investigate the invariance of face-selective units to face images of various sizes, positions, and rotation angles, the image set (Stigliani, 2015) was generated after modifying the size, position, and rotation angle of the faces and other objects in the low-level feature-controlled stimulus set.</p> <p>- <strong>IMG_var_view_210106.mat</strong> : This set was used to find units that invariantly responded to face images of different viewpoints. This dataset consists of five angle-based viewpoint classes (-90°, -45°, 0°, 45°, 90°) with ten different faces obtained from (Gourier 2004)</p> <p> </p> <p><strong>2. Data.zip</strong></p> <p>- <strong>Data_PFI_XDream/RevCorr_ClsUnit.mat</strong> : The obtained preferred feature images, using reverse-correlation method and X-Dream, of units selective to each object class.</p> <p>- <strong>Data_SVM_NeuronType.mat</strong> : Simulated face detection performance using distinct types of selective units in Conv5 and using the shuffled responses of face-selective units in untrained networks (a. all selective units, b. units selective to non-face classes (nonface-selective), c. face-selective units, d. units selective to none of these classes (non-selective)).</p> <p>- <strong>Data_SVM_invariance.mat</strong> : Face detection performance with variation of the low-level features. </p> <p>- <strong>PretrainedNet</strong> : Sample networks was untrained and trained with three types of image sets (a. face-reduced ImageNet, b. original ImageNet, c. original ImageNet with added face images (Stigliani, 2015)) (number of networks = 3). </p> <p> - Net_Untrained : untrained AlexNets<br> - Net_FD_ImageNet : AlexNets trained to face-reduced ImageNet<br> - Net_ImageNet : AlexNets trained to original ImageNet<br> - Net_ImageNetwFace : AlexNets trained to original ImageNet with added face images</p> <p>- <strong>Data_Trained.mat</strong> : Result summary of four types of networks above (number of networks = 10).</p> <p> </p>
Topographic deep neural networks predict the functional organization of the primate ventral visual pathway
<p>Recording of presentation at the Neuroscience 2021 annual meeting (held virtually). The abstract follows:</p> <p> </p> <p>The primate ventral visual pathway is organized into functional maps, including pinwheel-like arrangements of orientation-tuned neurons in primary visual cortex (V1) and patches of category-selective neurons in higher visual cortex. While deep convolutional neural networks (DCNNs) trained for object recognition accurately predict neural representations throughout the ventral pathway, they have no spatial layout for features at a given retinotopic location and are thus unable to predict the rich topographic organization of visual cortex. Here, we close this gap by first assigning each DCNN unit a position in a 2D cortical sheet, then training the network to minimize a cost function with two components: one encouraging accurate object recognition, and another favoring correlated responses among nearby units in each model layer (Figure 1A, 1B). </p> <p>We find that training with this composite spatial loss produces brain-like topographic maps in both early and later model layers (Figure 1B). Early layers contain smooth orientation preference maps with pinwheels, clusters of units preferring the same spatial frequency, and color-preference domains resembling V1 “blobs”. In a later layer of the same model, we observe clusters of category-selective units, e.g., face patches, whose spatial organization largely matches that found in primate higher visual cortex. Our model thus leverages local response correlations, which have been linked to theories of wire-length minimization, to accurately predict neuron responses and functional organization throughout the ventral visual pathway. In support of the wire-length minimization hypothesis, we find that our topographic DCNN would require shorter connections than a standard DCNN to support connections between similarly-tuned neurons within early (38% reduction) and later (31% reduction) model layers (Figure 1D). These results suggest that the functional organization of visual cortex can be explained by two constraints: the need to perform object recognition and pressure for local populations of neurons to have correlated responses.</p>
Automatic taxonomic identification based on the Fossil Image Dataset (>415,000 images) and deep convolutional neural networks
<p>The rapid and accurate taxonomic identification of fossils is of great significance in paleontology, biostratigraphy, and other fields. However, taxonomic identification is often labor-intensive and tedious, and the requisition of extensive prior knowledge about a taxonomic group also requires long-term training. Moreover, identification results are often inconsistent across researchers and communities. Accordingly, in this study, we used deep learning to support taxonomic identification. We used web crawlers to collect the Fossil Image Dataset (FID) via the Internet, obtaining 415,339 images belonging to 50 fossil clades. Then we trained three powerful convolutional neural networks on a high-performance workstation. The Inception ResNet v2 architecture achieved an average accuracy of 0.90 in the test dataset when transfer learning was applied. The clades of microfossils and vertebrate fossils exhibited the highest identification accuracies of 0.95 and 0.90, respectively. In contrast, clades of sponges, bryozoans, and trace fossils with various morphologies or with few samples in the dataset exhibited a performance below 0.80. Visual explanation methods further highlighted the discrepancies among different fossil clades and suggested similarities between the identifications made by machine classifiers and taxonomists. Collecting large paleontological datasets from various sources, such as the literature, digitization of dark data, citizen-science data, and public data from the Internet may further enhance deep learning methods and their adoption. Such developments will also possibly lead to image-based systematic taxonomy to be replaced by machine-aided classification in the future. Pioneering studies can include microfossils and some invertebrate fossils. To contribute to this development, we deployed our model on a server for public access at www.ai-fossil.com.</p>
Shared Data for De-scattering Deep Neural Network
<p>These images show mouse cortical layer 2/3 pyramidal neurons sparsely labeled with a cell fill (eYFP or mScarlet-I) to visualize the dendritic arbor, including dendritic spines. Cells were imaged in vivo using point-scan two-photon microscopy (PSTPM) and temporal focusing microscopy (TFM). These images were used for training and testing the De-Scattering Deep Neural Network (<a href="https://github.com/eleweiz/DeScattering_NN">https://github.com/eleweiz/DeScattering_NN</a>).</p>
A Probabilistic Framework for Mutation Testing in Deep Neural Networks - Models archive Part 2
<p>Models used as part of the paper "A Probabilistic Framework for Mutation Testing in Deep Neural<br> Networks ?" submitted to the journal Information and Software Technology</p> <p>Replication package using the data is available at https://github.com/FlowSs/PM</p>
A Probabilistic Framework for Mutation Testing in Deep Neural Networks - Models archive Part 3
<p>Models used as part of the paper "A Probabilistic Framework for Mutation Testing in Deep Neural<br> Networks ?" submitted to the journal Information and Software Technology</p> <p>Replication package using the data is available at https://github.com/FlowSs/PMT</p>
A Probabilistic Framework for Mutation Testing in Deep Neural Networks - Models archive Part 1
<p>Models used as part of the paper "A Probabilistic Framework for Mutation Testing in Deep Neural<br> Networks ?" submitted to the journal Information and Software Technology</p> <p>Replication package using the data is available at https://github.com/FlowSs/PMT</p>
Impact of training set size on the ability of deep neural networks to deal with label noise
<p>This is the data set accompanying the publication 'Impact of training set size on the ability of deep neural networks to deal with label noise'</p> <p> </p> <p><strong>License</strong></p> <p>This work is shared under the <em><code>Creative Commons Attribution ShareAlike 4.0 International License</code></em><code><em> </em>(CC BY-SA 4.0)</code><em>: </em>https://creativecommons.org/licenses/by-sa/4.0/</p> <p>The dataset is taken from the 6th SpaceNet challenge (https://spacenet.ai/sn6-challenge/) which also falls under the <code>CC BY-SA 4.0 license</code>. The original dataset was modified by removing objects from the buildings json files and converting the json files to png images.</p> <p>Creators of the SpaceNet dataset:</p> <p>Shermeyer, J., Hogan, D., Brown, J., Etten, A.V., Weir, N., Pacifici, F., Hänsch, R., Bastidas, A., Soenen, S., Bacastow, T.M., & Lewis, R.</p> <p>SpaceNet partners: https://spacenet.ai/about-us/</p> <p> </p>
Accelerating DNA-PAINT imaging with a deep neural network
<p>(<a href="https://zenodo.org/api/files/7030c526-533f-44ce-8b8c-01a979fd6b6f/RawFrames_P5-IS_20pM.tif?versionId=55a12821-8472-42f5-91b5-eabda1516e2b">RawFrames_P5-IS_20pM.tif</a>) Raw DNA-PAINT SMLM frames with isolated emitters taken on TOM20 labelled MNTB neuronal rat tissue with P5 imaging strand at 20 pM concentration. </p> <p>(<a href="https://zenodo.org/api/files/7030c526-533f-44ce-8b8c-01a979fd6b6f/Binned_HighDensity_30kpatches.tif?versionId=ac01dd00-3c44-4059-a1b1-3b1045fa636d">Binned_HighDensity_30kpatches.tif</a> and <a href="https://zenodo.org/api/files/7030c526-533f-44ce-8b8c-01a979fd6b6f/Binned_HighDensity_30kpatches.csv?versionId=31e792fc-fab5-43cd-af81-a271595586d0">Binned_HighDensity_30kpatches.csv</a>) Artificially summed high-density emitter patches with the corresponding emitter coordinates used for training the DeepSTORM neural network. </p> <p>(DeepSTORM_model_metadata.mat and DeepSTORM_model_weights_best.hdf5) The trained model metadata and weights used for all predicted images in the study.</p> <p>(Figure 4_Large_super-resolution_image.png) The large super-resolution image in Figure 4 and Figure S4.</p> <p>(High-density-frames_Images 1 - 5 .tif) Five high-emitter density raw frames for alpha-tubulin and TOM20 of 400 frames each.</p> <p>(<a href="https://zenodo.org/api/files/b98093d2-f689-4fef-8222-220c07ca4fe9/Ground-truth-rendered_Image4_Tubulin.tif">Ground-truth-rendered_Image 1 - 5</a>.tif) Five ground truth images rendered in Picasso (drift-corrected, linked localisations, pixel size 13.37 nm/pixel) for alpha-tubulin and TOM20.</p> <p>(<a href="https://zenodo.org/api/files/e726ff9b-45ba-4d3a-b3ba-3e6eaca03d35/Low%20density%20frames%20for%20GT%20images.zip">Low density frames for GT images.zip</a>) Five low-emitter density (0.5 nM) raw frames for alpha-tubulin and TOM20 of 10000 frames each which were used to render ground truth images.</p> <p>(<a href="https://zenodo.org/api/files/e726ff9b-45ba-4d3a-b3ba-3e6eaca03d35/Bassoon_Homer_datasets.zip">Bassoon_Homer_datasets.zip</a>) Three Bassoon and Homer datasets each with a ground truth image and the corresponding high-density raw frames (5 nM, 800 frames).</p>
Dataset from: Selecting deep neural networks that yield consistent attribution-based interpretations for genomics
<p>Deep neural networks (DNNs) have demonstrated great promise at taking DNA sequences as input and predicting a wide variety of functional activity. Post hoc attribution analysis has been employed to provide insights into the features learned by DNNs, often revealing patterns such as known motifs. However, attribution maps are noisy in practice to an extent that varies from model to model, even across DNNs that yield similar generalization performance. This makes it challenging to identify which high-performing DNN will provide trustworthy explanations. Here we propose a summary statistic that characterizes the consistency of learned features across a population of attribution maps which can be utilized as an additional criterion for model selection. We demonstrate the efficacy of this approach quantitatively using synthetic data and qualitatively with chromatin accessibility data. Together, this work advances our ability to select optimal DNNs that not only yield high generalization performance but also reliable attribution maps that will, in turn, accelerate scientific discovery in genomics.</p>
Data for "Efficient neural decoding of self-location with a deep recurrent network"
<p>Data for reproducing results with Bayesian decoders (MLE and Bayesian with memory) reported in the article</p> <p>"Efficient neural decoding of self-location with a deep recurrent network".</p> <p> </p> <p>This data should be used with the code found in https://github.com/NeuroCSUT/RatGPS and should be placed in the Bayesian/Data folder of the codebase.</p> <p> </p>
Data for: Scalable Flood Level Trend Monitoring with Surveillance Cameras using a Deep Convolutional Neural Network [Version 2]
<p>This package provides material that can be openly published for the paper "Scalable Flood Level Trend Monitoring with Surveillance Cameras using a Deep Convolutional Neural Network". It consists in the code used to generate results and figures as well as the weights of the deep convolutional neural networks trained to segment water in the surveillance camera images.</p>
Dataset for the article "A deep equivariant neural network approach for efficient hybrid density functional calculations"
<p>Dataset files for DeepH-hybrid, containing structures and preprocessed Hamiltonian matrices computed with HSE06 hybrid functional and ABACUS DFT package.</p> <p>Enclosed include:</p> <ol> <li>The four datasets corresponding to the four DeepH-hybrid models mentioned in DeepH-hybrid's paper, each containing material structures and electronic structure properties of </li> <li>Basis set files for the ABACUS DFT package</li> <li>Snapshot of a version of the additional codes utilized to preprocess hybrid DFT Hamiltonians (also available via this <a href="https://github.com/aaaashanghai/DeepH-hybrid">GitHub link</a>)</li> </ol> <p>Note only the additional codes of DeepH-hybrid is provided, and it may be used in combination with DeepH-E3 package for neural-network training. For additional details please refer to the "README" file of the code.</p>
Characterization of deep neural network features by decodability from human brain activity
<p>We present a dataset derived through the DNN feature decoding analyses (<a href="https://www.nature.com/articles/ncomms15037">Horikawa and Kamitani, 2017</a>), including true and decoded feature values of DNNs (AlexNet and VGG19) and decoding accuracies of individual DNN features with their rankings. The decoding accuracies of individual DNN features were highly correlated across subjects, suggesting the systematic differences between the brain and DNNs. The unpreprocessed fMRI data is available from the OpenNeuro (<a href="https://openneuro.org/datasets/ds001246">https://openneuro.org/datasets/ds001246</a>). We hope the present dataset will contribute to reveal the gap between the brain and DNNs and provide an opportunity to make use of the decoded features for further applications.</p>
Convolution, aggregation and attention based deep neural networks for accelerating simulations in mechanics [Dataset]
<p>Supplementary data for 'Convolution, aggregation and attention based deep neural networks for accelerating simulations in mechanics'. </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.