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.
173
datasets available to search
ShareScore release 0.9.0
Dataset results
173 results for “convolutional neural network”
Training and validation datasets for "Three-Dimensional Implicit Structural Modeling Using Convolutional Neural Network"
<p>This is training and validation datasets used in manuscript "Three-Dimensional Implicit Structural Modeling Using Convolutional Neural Network". In this manuscript, we propose an efficient deep learning method using a Convolutional Neural Network (CNN) to predict a scalar field from sparse structural data associated with multiple distinct stratigraphic layers and faults. The CNN architecture is beneficial for the flexible incorporation of empirical geological knowledge when trained with numerous and realistic structural models that are automatically generated from a data simulation workflow. It also presents an expressive characteristic of integrating various types of structural constraints by optimally minimizing a hybrid loss function to compare predicted and reference structural models, opening new opportunities for further improving geological modeling. </p>
Training and validation data for artificial neural networks using three-dimensional partial convolutions to fill gaps in satellite image time series
<p>This dataset contains training and validation data for artificial neural networks using three-dimensional partial convolutions to fill gaps in satellite image time series. The data have been derived from Sentinel-5P total column carbon monoxide observations, using the offline processing stream.</p> <p><strong>Preprocessing</strong></p> <p>The following operations have been applied on the original S5P imagery:</p> <ol> <li>Images have been resampled to 0.1 by 0.1 degree spatial resolution</li> <li>Pixels with quality assessment value less than or equal to 0.5 have been set to NA</li> <li>Images have been aggregated by day of observation</li> <li>Images have been cropped to -60 to 60 degrees latitude</li> <li>Images have been devided into spatiotemporal blocks of size 128 x 128 pixels and 16 days</li> </ol> <p>Imagery has been recorded between 2021-01-01 and 2021-11-25. Notice that both the training and the validation blocks have been randomly sampled from all available blocks.</p> <p><br> <strong>Data Format and Naming Conventions</strong></p> <p>Input and output data blocks are stored as GeoTIFF files, where bands represent time. Notice the following file naming conventions:</p> <ul> <li>Files starting with <em>X</em> represent input measurements for training, where artificial gaps have been added.</li> <li>Files starting with <em>Y</em> represent true measurements without artificially added gaps (but still containing gaps in many cases).</li> <li>Binary masks of input data where all pixels with valid measurements are 1 and others 0 are stored in files whose name starts with <em>MASK</em></li> <li>Files starting with <em>VALMASK</em> contain a binary mask where only pixels that are available in Y but not in X are 1. The latter is used for validation on artificially removed pixels only.</li> </ul> <p>Numbers in filenames encode spatial and temporal block indexes.</p> <p>In addition, the dataset contains prediction of the validation blocks from different models in the `predictions` directory. The subfolders contain output from different models:</p> <ul> <li>mean refers to simple block-wise mean predictions.</li> <li>timeseries refers to simple linear time series interpolation.</li> <li>gapfill refers to the method proposed in [1].</li> <li>stmra refers to the method proposed in [2].</li> <li>STpconv refers to predictions passed on an artificial neural netowork with three-dimensional partial convolutions.</li> </ul> <p><strong>References</strong></p> <p>[1] Gerber, F., de Jong, R., Schaepman, M. E., Schaepman-Strub, G., & Furrer, R. (2018). Predicting missing values in spatio-temporal remote sensing data. IEEE Transactions on Geoscience and Remote Sensing, 56(5), 2841-2853.</p> <p>[2] Appel, M., & Pebesma, E. (2020). Spatiotemporal multi-resolution approximations for analyzing global environmental data. Spatial Statistics, 38, 100465.</p>
Using deep convolutional neural networks to forecast spatial patterns of Amazonian deforestation: supporting data and outputs
<p class="MsoNormal"><strong>1. </strong>Tropical forests are subject to diverse deforestation pressures while their conservation is essential to achieve global climate goals. Predicting the location of deforestation is challenging due to the complexity of the natural and human systems involved but accurate and timely forecasts could enable effective planning and on-the-ground enforcement practices to curb deforestation rates. New computer vision technologies based on deep learning can be applied to the increasing volume of Earth observation data to generate novel insights and make predictions with unprecedented accuracy.</p> <p class="MsoNormal"><strong>2. </strong>Here, we demonstrate the ability of deep convolutional neural networks (CNNs) to learn spatiotemporal patterns of deforestation from a limited set of freely available global data layers, including multispectral satellite imagery, the Hansen maps of annual forest change (2001-2020) and the ALOS PALSAR digital surface model, to forecast deforestation (2021). We designed four model architectures, based on 2D CNNs, 3D CNNs, and Convolutional Long Short-Term Memory (ConvLSTM) Recurrent Neural Networks (RNNs), to produce spatial maps that indicate the risk to each forested pixel (~30 m) in the landscape of becoming deforested within the next year. They were trained and tested on data from two ~80,000 km<sup>2</sup> tropical forest regions in the Southern Peruvian Amazon.</p> <p class="MsoNormal"><strong>3.</strong><strong> </strong><span>The networks could predict the location of future forest loss to a high degree of accuracy (F</span><sub>1 </sub><span>= 0.58-0.71). Our best performing model (3D CNN) had the highest pixel-wise accuracy (F</span><sub>1 </sub><span>= 0.71) when validated on 2020 forest loss (2014-2019 training). Visual interpretation of the mapped forecasts indicated that the network could automatically discern the drivers of forest loss from the input data. For example, pixels around new access routes (e.g. roads) were assigned high risk whereas this was not the case for recent, concentrated natural loss events (e.g. remote landslides).</span></p> <p class="MsoNormal"><strong>4.</strong><strong> </strong>CNNs can harness limited time-series data to predict near-future deforestation patterns, an important step in harnessing the growing volume of satellite remote sensing data to curb global deforestation. The modelling framework can be readily applied to any tropical forest location and used by governments and conservation organisations to prevent deforestation and plan protected areas.</p>
Classification of tropical cyclone containing images using a convolutional neural network: performance and sensitivity to the learning dataset
<p>NXTensor extraction library, experiment code, tropical cyclone and background images and their metadata generated from the meterological reanalysis ERA5 and MERRA-2 according to the HURDAT2 cyclone tracks.</p> <p>Version specifications:</p> <ul> <li>NXTensor: v0.3.3.10</li> <li>Experiment code: v2.0.3</li> <li>Image sets: v1</li> </ul> <p> </p>
Supplemental material for 'Characterization of structure and mixing in nanoparticle hetero-aggregates using convolutional neural networks: 3D-reconstruction versus 2D-projection'
<p>This is the supplemental data for the manuscript titled ‘<em>Characterization of structure and mixing in nanoparticle hetero-aggregates using convolutional neural networks: 3D-reconstruction versus 2D-projection’</em> submitted to <em>Ultramicroscopy</em>.</p> <p><strong>Motivation:</strong></p> <p>Detection of nanoparticles and classification of the material type in scanning transmission electron microscopy (STEM) images can be a tedious task, if it has to be done manually. Therefore, a convolutional neural network (CNN) is trained to do this task for STEM-images of TiO<sub>2</sub>-WO<sub>3</sub> nanoparticle hetero-aggregates. In conventional STEM, only 2D-projection images of the samples can be measured. STEM tomography allows for a 3D-reconstruction but it is a time-consuming and hence expensive task. In the present work, evaluations of 2D-projections are compared quantitatively to evaluations of 3D-reconstructions. For both evaluations a CNN is trained to predict particle positions and classify the material. The present dataset contains training and evaluation data and some code scripts that can be used after installation of the MMDetection toolbox (<a href="https://github.com/open-mmlab/mmdetection">https://github.com/open-mmlab/mmdetection</a>) to train the CNN for 2D-projection data. For 3D-reconstruction a StarDist-3D network is trained (<a href="https://github.com/stardist/stardist">https://github.com/stardist/stardist</a>). Details are provided in the manuscript submitted to Ultramicroscopy and in the comments of the code scripts. For evaluation, we provide Python and MATLAB scripts.</p> <p><strong>Authors and funding:</strong></p> <p>The present dataset was created by the authors. The work was funded by the Deutsche Forschungsgemeinschaft within the priority program SPP2289 under contract numbers RO2057/17-1 and MA3333/25-1 and under contract number INST 144/462-1 FUGG.</p> <pre> </pre> <p><strong>Dataset description:</strong></p> <p>We provide several zip-archives. All of them contain two subfolders, one of them for 2D-projection data, the other one for 3D-reconstruction data.</p> <p><em>training_data.zip</em> contains the training data. In the 3D case, the subfolder <em>mask</em> contains the ground truth segmentation masks; the subfolder <em>reconstruction</em> contains the corresponding simulated 3D reconstructions. In the 2D case, the subfolder <em>HAADF</em> contains the 2D-projection images. In both cases, the subfolder <em>json </em>contains the annotation. Each file within the <em>json</em> folder provides for each image or reconstruction the following information:</p> <p>· aggregat_no: image id, the number of the corresponding image file</p> <p>· particle_position_x: list of particle position x-coordinates in nm</p> <p>· particle_position_y: list of particle position y-coordinates in nm</p> <p>· particle_position_z: list of particle position z-coordinates in nm</p> <p>· particle_radius: list of volume equivalent particle radii in nm</p> <p>· particle_type: list of particle types, 1: TiO<sub>2</sub>, 2: WO<sub>3</sub></p> <p>· particle_shape: list of particle shapes: 0: sphere, 1: box, 2: icosahedron</p> <p>· rotation: list of particle rotations in rad. Each particle is rotated twice by the listed angle (before and after deformation)</p> <p>· deformation: list of particle deformations. After the first rotation the particle x-coordinates of the particle’s surface mesh are scaled by the factor listed in deformation, y- and z-coordinates are scaled according to 1/sqrt(deformation).</p> <p>· cluster_index: list of cluster indices for each particle</p> <p>· initial_cluster_index: list of initial cluster indices for each particle, before primary clusters of the same material were merged</p> <p>· fractal_dimension: the intended fractal dimension of the aggregate</p> <p>· fractal_dimension_true: the realized geometric fractal dimension of the aggregate (neglecting particle densities)</p> <p>· fractal_dimension_weight_true: the realized fractal dimension of the aggregate (including particle densities)</p> <p>· fractal_prefactor: fractal prefactor</p> <p>· mixing_ratio_intended: the intended mixing ratio (fraction of WO<sub>3</sub> particles)</p> <p>· mixing_ratio_true: the realised mixing ratio (fraction of WO<sub>3</sub> particles)</p> <p>· mixing_ratio_volume: the realised mixing ratio (fraction of WO<sub>3</sub> volume)</p> <p>· mixing_ratio_weight: the realised mixing ratio (fraction of WO<sub>3</sub> weight)</p> <p>· particle_1_rho: density of TiO<sub>2</sub> used for the calculations</p> <p>· particle_1_size_mean: mean TiO<sub>2</sub> radius</p> <p>· particle_1_size_min: smallest TiO<sub>2</sub> radius</p> <p>· particle_1_size_max: largest TiO<sub>2</sub> radius</p> <p>· particle_1_size_std: standard deviation of TiO<sub>2</sub> radii</p> <p>· particle_1_clustersize: average TiO<sub>2</sub> cluster size</p> <p>· particle_1_clustersize_init: average TiO<sub>2</sub> cluster size of primary clusters (before merging into larger clusters)</p> <p>· particle_1_clustersize_init_intended: intended TiO<sub>2</sub> cluster size of primary clusters</p> <p>· particle_2_rho: density of WO<sub>3 </sub>used for the calculations</p> <p>· particle_2_size_mean: mean WO<sub>3</sub> radius</p> <p>· particle_2_size_min: smallest WO<sub>3</sub> radius</p> <p>· particle_2_size_max: largest WO<sub>3</sub> radius</p> <p>· particle_2_size_std: standard deviation of WO<sub>3</sub> radii</p> <p>· particle_2_clustersize: average WO<sub>3</sub> cluster size</p> <p>· particle_2_clustersize_init: average WO<sub>3</sub> cluster size of primary clusters (before merging into larger clusters)</p> <p>· particle_2_clustersize_init_intended: intended WO<sub>3</sub> cluster size of primary clusters</p> <p>· number_of_primary_particles: number of particles within the aggregate</p> <p>· gyration_radius_geometric: gyration radius of the aggregate (neglecting particle densities)</p> <p>· gyration_radius_weighted: gyration radius of the aggregate (including particle densities)</p> <p>· mean_coordination: mean total coordination number (particle contacts)</p> <p>· mean_coordination_heterogen: mean heterogeneous coordination number (contacts with particles of the different material)</p> <p>· mean_coordination_homogen: mean homogeneous coordination number (contacts with particles of the same material)</p> <p>· material_1: the name of the first material (TiO2)</p> <p>· material_2: the name of the second material (WO3)</p> <p>· radius_equiv: list of area equivalent particle radii (in projection) in nm</p> <p>· k_proj: projection direction of the aggregate: 0: z-direction (axis = 2), 1: x-direction (axis = 1), 2: y-direction (axis = 0)</p> <p>· polygons: list of polygons that surround the particle (COCO annotation)</p> <p>· bboxes: list of particle bounding boxes</p> <p>· aggregate_size: projected area of the aggregate translated into the radius of a circle in nm</p> <p>· n_pix: number of pixel per image in horizontal and vertical direction (squared images)</p> <p>· pixel_size: pixel size in nm</p> <p>· image_size: image size in nm</p> <p>· add_poisson_noise: 1 if poisson noise was added, 0 otherwise</p> <p>· frame_time: simulated frame time (required for poisson noise)</p> <p>· dwell_time: dwell time per pixel (required for poisson noise)</p> <p>· beam_current: beam current (required for poisson noise)</p> <p>· electrons_per_pixel: number of electrons per pixel</p> <p>· dose: electron dose in electrons per Å<sup>2</sup></p> <p>· add_scan_noise: 1 if scan noise was added, 0 otherwise</p> <p>· beam_misposition: parameter that describes how far the beam can be misplaced in pm (required for scan noise)</p> <p>· scan_noise: parameter that describes how far the beam can be misplaced in pix (required for scan noise)</p> <p>· add_focus_dependence: 1 if a focus effect is included, 0 otherwise</p> <p>· add_partial_coating: 1 if some TiO<sub>2</sub> particles were coated by a thin WO<sub>3</sub> film, 0 otherwise.</p> <p>· data_format: data format of the images, e.g. uint8</p> <p>For 3D reconstructions, the following information is added:</p> <p>· add_image_shifts, 1 if all images of the tilt series were shifted randomly by some pixel to account for misaligned images, 0 otherwise.</p> <p>· add_projection_noise: 1 if noise was added to projection angles, 0 otherwise.</p> <p>· N_SIRT: number of SIRT iterations for the 3D-reconstruction.</p> <p>· proj_angles: list of angles used for the projection directions of the tilt series in rad.</p> <p>For the 2D case, there are 24000 training images, 5500 validation images, 5500 test images, and their corresponding annotations. Aggregates and STEM images were obtained with the algorithm explained in the main work. The important data for CNN training is extracted from the files of individual aggregates and concluded in the subfolder <em>COCO</em>. For training, validation and test data there is a file <em>annotation_COCO.json</em> that includes all information required for the CNN training.</p> <p>For the 3D case, there are 100 simulated reconstructions that are divided into 80 training and 20 validation images within the training script.</p> <p>The zip archive <em>models.zip</em> includes the two networks that were trained, evaluated and used for the investigation in the manuscript. In the 2D case, network weights are stored in the file <em>2D_projection/logs/fit/20240209-095952/iter_60000.pth</em>. These weights can be loaded with the jupyter-notebook <em>2D_projection_prediction.ipynb</em>. Furthermore, a configuration file, which is required by the notebooks, is stored as <em>2D_projection/logs/fit</em> <em>20240209-095952/config_file.py</em>. In the 3D case, network weights are stored in <em>3D_reconstruction/model/weights_best.h5</em>. Also for this case, a configuration file is provided. The network can be loaded with the jupyter-notebook <em>3D_reconstruction_prediction.ipynb</em>.</p> <p>The zip archive <em>experiment_measurement.zip</em> includes the experimental 2D-projection images and the experimental 3D-reconstructions investigated in the manuscript. In the 3D case, we provide measured reconstructions as obtained by MATLAB and as transformed for the prediction with the CNN.</p> <p>The zip archive<em> experiment_prediction.zip</em> includes predictions and visualizations obtained by the 2D and 3D CNNs.</p> <p>The zip archive <em>simulation_measurement.zip</em> includes simulated 2D-projection images and simulated 3D-reconstructions that were not used for the network training. These simulations were used for evaluation of trained networks.</p> <p>The zip archive<em> simulation_prediction.zip</em> includes predictions and visualizations obtained by the 2D and 3D CNNs for the simulated data that was not used during the training process.</p> <p>In the zip archive <em>code.zip</em>, we provide several files with Python and MATLAB code that can be used for training, prediction and evaluation of the CNNs. A lot of information is provided within the comments and markdowns. For the application, it is required that the MMDetection toolbox and the StarDist3D framework are installed.</p> <p>· <em>2D_projection_training.py:</em> This Python script can be used for network training of the 2D Mask R-CNN after installation of the MMDetection toolbox.</p> <p>· <em>3D_reconstruction_training.py</em>: This Python script can be used for network training of the StarDist-3D network after installation of the StarDist package.</p> <p>· <em>2D_projection_prediction.ipynb</em>: This jupyter-notebook can be used for the application of a trained Mask R-CNN to experimental and simulated 2D-projection data.</p> <p>· <em>3D_reconstruction_prediction.ipynb</em>: This jupyter-notebook can be used for the application of a trained StarDist-3D network to experimental and simulated 3D-reconstruction data.</p> <p>· <em>Evaluation_experiment.m</em>: This MATLAB script is for the visualization and quantitative comparison of 2D and 3D experimental evaluations.</p> <p>· <em>Evaluation_simulation.m</em>: This MATLAB script is for the visualization and quantitative comparison of 2D and 3D evaluations of simulations.</p> <p>· <em>particle_detection_functions.py:</em> This Python script contains functions required by the jupyter-notebooks. Details can be found within the comments.</p> <p>· <em>ASTRA_CM_plot_results.m</em>: This MATLAB script provides functions for the visualization of experimental and simulated 3D-reconstructions.</p> <p>· <em>ASTRA_CM_particle_detection.m</em>: This MATLAB script is used for the quantitative evaluation of segmentations of 3D-reconstructions.</p> <p> </p> <p>There is no confidential data in this dataset. It is neither offensive, nor insulting or threatening.</p> <p>The dataset was generated to discriminate between TiO<sub>2 </sub>and WO<sub>3</sub> nanoparticles in STEM-images and STEM tomography reconstructions. It might be possible that it can discriminate between different materials if the STEM contrast is similar to the contrast of TiO<sub>2 </sub>and WO<sub>3</sub> but there is no guarantee.</p>
Dataset for: Asphalt pavement crack detection based on convolutional neural network and infrared thermography
<p>This is the dataset for the following paper: </p> <p>Fangyu Liu, Jian Liu, and Linbing Wang. "Asphalt pavement crack detection based on convolutional neural network and infrared thermography." IEEE Transactions on Intelligent Transportation Systems 23, no. 11 (2022): 22145-22155. https://doi.org/10.1109/TITS.2022.3142393. </p> <p>Data component:</p> <ul> <li>01-Visible images: this folder includes fully visible images</li> <li>02-Infrared images: this folder includes fully infrared images</li> <li>03-Fusion(50IRT) images: this folder includes fusion images (50% infrared + 50% visible)</li> <li>04-Ground truth: this folder includes ground truth (binary images)</li> </ul>
Deep convolutional neural network for owl vocal identification
<p>This repository contains all the code and data necessary to replicate the results presented in Ruff et al. 2019, "Automated identification of avian vocalizations with deep convolutional neural networks", and is published in support of that manuscript. The folder includes several Python scripts, our trained convolutional neural network (CNN), and a set of 164,210 spectrogram images that were reviewed to generate CNN performance metrics. We include the CNN's predicted class scores for the test images as well as the set of labels assigned to the same images by experienced human technicians. The published article can be found here: <a href="https://zslpublications.onlinelibrary.wiley.com/doi/full/10.1002/rse2.125">https://zslpublications.onlinelibrary.wiley.com/doi/full/10.1002/rse2.125</a></p> <p>As presented, the CNN is designed to accept grayscale PNG images at 500x129 resolution and will generate a set of seven class scores for each image. Class scores are the softmax activation from the final (seven unit) fully-connected layer of the CNN. Scores are bounded between 0 and 1 and sum to 1 for each image. This means target classes are implicitly treated as mutually exclusive (i.e., each image belongs to exactly one class), although in reality some images contain calls from >1 target species.</p> <p>The different scripts and their functions are as follows:<br> - Code used to construct and train the CNN is in Owl_CNN_train_model.py<br> - Code to generate spectrograms with randomized parameters based on tagged calls in audio files is in Owl_CNN_generate_training_data.py<br> - Code to generate random spectrograms from a set of audio files (used to generate training data for the Noise class) can be generated with Owl_CNN_make_noise_data.py<br> - Code used to process raw audio files, including segmenting them into 12 s clips, generating spectrograms, and generating class scores using a pre-trained CNN is in Owl_CNN_process_audio.py<br> - Code to generate class scores for an existing set of spectrogram images using a pre-trained CNN are in Owl_CNN_process_images.py</p> <p>Our seven target classes are as follows:<br> AEAC - Northern saw-whet owl, Aegolius acadicus.<br> BUVI - Great horned owl, Bubo virginianus.<br> GLGN - Northern pygmy-owl, Glaucidium gnoma.<br> MEKE - Western screech-owl, Megascops kennicottii.<br> STOC - (Northern) spotted owl, Strix occidentalis caurina.<br> STVA - Barred owl, Strix varia.<br> Noise - Catch-all for any clip that did not contain vocalizations of at least one of the six owl species listed above.</p> <p>The CNN was trained for 100 epochs and saved only after epochs in which validation loss improved. Loss was measured as categorical cross-entropy. The CNN was last saved at epoch 97 with reported metrics:<br> Training loss = 0.218<br> Training accuracy = 0.972<br> Validation loss = 0.165<br> Validation accuracy = 0.987</p> <p>Although this code has been tested and works on our system, we make no guarantee that it will work for others without modification. Created using Python version 2.7.14, TensorFlow version 1.2.1, Keras version 2.2, and SoX version 14.4. Code was developed by Bharath Padmaraju, Zack Ruff, and Chris Sullivan. Questions and comments may be directed to zjruff at gmail dot com.</p> <p>Zack Ruff<br> 15 July 2019</p>
Channel State Information (CSI) analysis for predictive maintenance using Convolutional Neural Network (CNN)
<p>Dataset manual:</p> <p>This dataset contains CSI amplitude values for rotating motors in an office environment. Details of the experiments may be found in the corresponding paper published in the DATA'19 workshop, SenSys (<a href="https://doi.org/10.1145/3359427.3361917">https://doi.org/10.1145/3359427.3361917</a>). </p> <p>Folder structure:<br> The folders for servo motor and stepper motor contains separate folders for network reconnection conditions (w_recc: with reconnections, wo_recc: without recconnections) and load conditions (w_load: with load and wo_load: without load). The data is stores as Matlab files with .mat extentions. </p> <p>File structure:<br> In each file name, the digits after the '_' at the end of the file name correspond to the speed of the motor. In case of stepper motor these numbers could be directly interpreted as rpm. Ex: table_inj_with_load_5_0.mat corresponds to stationary motor (0 rpm) and table_inj_with_load_5_250.mat corresponds to motor rotating with 250 rpm speed. In the case of servo motor these numbers should be mapped with the following table in order to get the speeds.</p> <p>0: 0 rpm<br> 50: 14.45 rpm<br> 100: 8.02 rpm<br> 150: 5.38 rpm<br> 200: 4.05 rpm<br> 250: 3.26 rpm<br> 300: 2.67 rpm<br> Ex: table_inj_with_load_50.mat corresponds to motor running with 14.45 rpm.</p> <p>Each file has 3 columns, each corresponding to CSI value, labels (speed/last digits in the file name) and the data sample number (not in sequence as a result of packer loss) respectively. CSI values are typically a matrix of size 3000*180 (3000 CSI samples for 3sec data @1kHz sampling rate and 180 channels for 6 antenna pairs @ 30 subcarrier data per antenna).</p>
Convolutional Neural Networks for Classification of Alzheimer's Disease: Overview and Reproducible Evaluation [Models]
<p>This file contains the pretrained models and the evaluation of the pipelines described in the paper <em>Convolutional Neural Networks for Classification of Alzheimer’s Disease: Overview and Reproducible Evaluation</em>.</p> <p>Source code can be downloaded at: <a href="https://github.com/aramis-lab/AD-DL">https://github.com/aramis-lab/AD-DL</a></p> <p>Also, single files can be obtained at: <a href="https://aramislab.paris.inria.fr/clinicadl/files/models/v0.0.1/">https://aramislab.paris.inria.fr/clinicadl/files/models/v0.0.1/</a></p> <p>The structure of the compressed file is as follows:</p> <p>clinicadl_models/<br> ├── 2D_slice<br> │ ├── baseline<br> │ │ ├── AD_CN<br> │ │ │ ├── best_model<br> │ │ │ └── performances<br> │ │ └── AD_CN_dataleakage<br> │ │ ├── best_model<br> │ │ └── performances<br> │ └── longitudinal<br> │ └── AD_CN<br> │ ├── best_model<br> │ └── performances<br> ├── 3D_patch<br> │ ├── baseline<br> │ │ ├── AD_CN<br> │ │ │ ├── best_model<br> │ │ │ └── performances<br> │ │ └── sMCI_pMCI<br> │ │ ├── best_model<br> │ │ └── performances<br> │ └── longitudinal<br> │ ├── AD_CN<br> │ │ ├── best_model<br> │ │ └── performances<br> │ └── sMCI_pMCI<br> │ ├── best_model<br> │ └── performances<br> ├── 3D_ROI_based<br> │ ├── baseline<br> │ │ ├── AD_CN<br> │ │ │ ├── best_model<br> │ │ │ └── performances<br> │ │ └── sMCI_pMCI<br> │ │ ├── best_model<br> │ │ └── performances<br> │ └── longitudinal<br> │ ├── AD_CN<br> │ │ ├── best_model<br> │ │ └── performances<br> │ └── sMCI_pMCI<br> │ ├── best_model<br> │ └── performances<br> ├── 3D_subject<br> │ ├── baseline<br> │ │ ├── AD_CN<br> │ │ │ ├── best_model<br> │ │ │ └── performances<br> │ │ └── sMCI_pMCI<br> │ │ ├── best_model<br> │ │ └── performances<br> │ └── longitudinal<br> │ ├── AD_CN<br> │ │ ├── best_model<br> │ │ └── performances<br> │ └── sMCI_pMCI<br> │ ├── best_model<br> │ └── performances<br> ├── autoencoders<br> │ ├── 3D_patch<br> │ │ ├── baseline<br> │ │ │ └── best_model<br> │ │ └── longitudinal<br> │ │ └── best_model<br> │ ├── 3D_ROI_based<br> │ │ ├── baseline<br> │ │ │ └── best_model<br> │ │ └── longitudinal<br> │ │ └── best_model<br> │ └── 3D_subject<br> │ └── baseline<br> │ ├── extensive<br> │ └── minimal<br> └── svm<br> ├── baseline<br> │ ├── AD_CN<br> │ │ ├── all_subjects.tsv<br> │ │ └── classifier<br> │ └── sMCI_pMCI<br> │ ├── all_subjects.tsv<br> │ └── classifier<br> └── longitudinal<br> ├── AD_CN<br> │ ├── all_subjects.tsv<br> │ └── classifier<br> └── sMCI_pMCI<br> ├── all_subjects.tsv<br> └── classifier</p> <p>We provide the pretrained CNN models for the frameworks 3D subject-level, 3D ROI-based, 3D patch-level and 2D slice-level. This models can be found as a <strong><em>.pth.tar</em> </strong>file (<em>Pytorch</em> format) inside the <em>best_model</em> folder for each framework (and for each fold). We also provide the autoencoders that initialize the training stage of the CNN networks. The <em>performances </em>folder contains the computed metrics for the correponding model (ACC, BA, etc). <em> </em></p> <p>For the svn classification, we provide files with the dual coefficients, the support vector indices and the weights. Also, <em>tsv</em> files with the subject list.</p>
Data for: Redshift Prediction with Images for Cosmology using a Bayesian Convolutional Neural Network with Conformal Predictions
<p>These files contain the predictions from the CNN and BCNN model from the paper titled: "Redshift Prediction with Images for Cosmology using a Bayesian Convolutional Neural Network with Conformal Predictions" (Jones et al. 2024). These files will allow reproduction of the performance metrics described in the paper.</p> <p> </p> <p>full_prediction_set_CNN.csv - predictions for the redshift using the CNN model of the entire dataset<br>cnn_evaluation.csv - predictions from just the evaluation dataset that was not used in training</p> <p>Columns are:</p> <p>photoz - predicted photoz from the model<br>specz - spectroscopic redshift<br>objectid - object ID from HSC PDR2 data release (Aihara et al. 2019)</p> <p><br>full_prediction_set_BCNN.csv - predictions for the redshift using the BCNN model of the entire dataset<br>bcnn_evaluation.csv - predictions from just the evaluation dataset that was not used in training</p> <p>Columns are:</p> <p>photoz - predicted photoz from the model<br>specz - spectroscopic redshift<br>objectid - object ID from HSC PDR2 data release (Aihara et al. 2019)<br>photoz_uncertainty - uncertainty in the predicted photoz</p>
Datasets from "A new method for the detection of siliceous microfossils on sediment microscope slides using convolutional neural networks". JGR Biogeosciences.
<p>This repository contains the datasets linked to "A new method for the detection of siliceous microfossils on sediment microscope slides using convolutional neural networks" (Journal of Geophysical Research: Biogeosciences, <a href="https://doi.org/10.1029/2024JG008047">https://doi.org/10.1029/2024JG008047</a>). This includes:</p> <ul> <li>The images and annotation* text files (in YOLO format) used for the detection of siliceous microfossils. </li> <li>The models and training results for the trainings presented in the main text and supporting information.</li> </ul> <p>*Note that while annotations were attributed to 14 general categories, only twelve of these were used during training (<em>i.e. </em>Pennate, Centric, Silicoflagellate, Centric_debris, Spore, Other_Biomin, Cocco, Silicoflagellate_debris, Undetermined_silica, Chateoceros_Bacteriastrum, Calcispheres, Foraminifer), and all were pooled into a single "Microfossil" category for the purpose of the training.</p>
Deep learning based on convolutional neural networks to classify nanobiomechanical data
Open the record for dataset details and reuse information.
Pretraining convolutional neural networks for mudstones petrographic thin section image classification
<p>This dataset was used in the paper "Pretraining convolutional neural networks for mudstones petrographic thin section image classification"</p>
Dataset - seismic data from central-western Italy used in the paper on rapid prediction of ground motion using a Convolutional Neural Network
<p>The dataset published here is the central-western Italy dataset used in the paper "<em>Transfer learning: Improving neural network based prediction of earthquake ground shaking for an area with insufficient training data"</em> (<a href="https://arxiv.org/abs/2105.05075">https://arxiv.org/abs/2105.05075</a>). The code for the paper is available at <a href="https://github.com/djozinovi/TLpredIM">https://github.com/djozinovi/TLpredIM</a>. The abstract of the paper:</p> <blockquote> <p>In a recent study (Jozinović et al, 2020) we showed that convolutional neural networks (CNNs) applied to network seismic traces can be used for rapid prediction of earthquake peak ground motion intensity measures (IMs) at distant stations using only recordings from stations near the epicenter. The predictions are made without any previous knowledge concerning the earthquake location and magnitude. This approach differs from the standard procedure adopted by earthquake early warning systems (EEWSs) that rely on location and magnitude information. In the previous study, we used 10 s, raw, multistation waveforms for the 2016 earthquake sequence in central Italy for 915 events (CI dataset). The CI dataset has a large number of spatially concentrated earthquakes and a dense station network. In this work, we applied the CNN model to an area around area near Pisa, Italy. In our initial application of the technique, we used a dataset consisting of 266 earthquakes recorded by 39 stations. We found that the CNN model trained using this smaller dataset performed worse compared to the results presented in the original study by Jozinović et al. (2020). To counter the lack of data, we adopted transfer learning (TL) using two approaches: first, by using a pre-trained model built on the CI dataset and, next, by using a pre-trained model built on a different (seismological) problem that has a larger dataset available for training. We show that the use of TL improves the results in terms of outliers, bias, and variability of the residuals between predicted and true IMs values. We also demonstrate that adding knowledge of station positions as an additional layer in the neural network improves the results. The possible use for EEW is demonstrated by the times for the warnings that would be received at the station PII.</p> </blockquote>
Test dataset for "Rapid estimation of cortical neuron activation thresholds by transcranial magnetic stimulation using convolutional neural networks"
<p>Data corresponding to test dataset used in Aberra AS, Lopez A, Grill WM, Peterchev AV. (2022). "Rapid estimation of cortical neuron activation thresholds by transcranial magnetic stimulation using convolutional neural networks". bioRxiv. Dataset includes:</p> <ul> <li><em>simnibs/ -</em> SimNIBS mesh and E-field solution file used in test dataset (posterior-anterior TMS of M1 in <em>ernie</em> example mesh, meshed with mri2mesh pipeline)</li> <li><em>layer_data/ - </em>surface meshes used for placing and orienting neuron models and corresponding sampling grids for CNNs</li> <li><em>nrn_sim_data/ - </em>Thresholds from NEURON simulations for all 25 model neurons included in the study, each at 4,999-5,000 positions and 12 azimuthal orientations ("ground truth" for CNN) </li> <li><em>cell_data/</em> - Coordinates and morphology information for all model neurons</li> <li><em>weights/</em> - Trained 3D convolutional neural networks for estimating neuron model-specific TMS thresholds given input E-field distributions on a 3D grid (see code/manuscript for dimensions)</li> <li><em>est_data/ </em>- Output of trained CNNs on all E-field data for test dataset <em> </em></li> </ul> <p> </p>
Data for: Reconstructing Cosmological Initial Conditions from Late-Time Structure with Convolutional Neural Networks
<p>Trained models and evaluation data for the revised submitted paper "Reconstructing Cosmological Initial Conditions from Late-Time Structure with Convolutional Neural Networks," Christopher J. Shallue & Daniel J. Eisenstein (2022)</p>
Supplementary Data for "Prediction of solar wind speed by applying convolutional neural network to potential field source surface (PFSS) magnetograms"
<p>These are supplementary data for the paper "Prediction of solar wind speed by applying convolutional neural network to potential field source surface (PFSS) magnetograms". They are:</p> <p>- Python code to construct a neural network model</p> <p>- Saved optimal models (for 8-fold validation)</p> <p>- Selected y-label data (solar wind speed) and corresponding dates, which we eliminate the data identified as ICME</p>
Supplemental data for characterization of mixing in nanoparticle hetero-aggregates using convolutional neural networks
<p>This is the supplemental data for the manuscript titled <em>Characterization of mixing in nanoparticle hetero-aggregates using convolutional neural networks</em> submitted to <em>Nano Select</em>.</p> <p><strong>Motivation:</strong></p> <p>Detection of nanoparticles and classification of the material type in scanning transmission electron microscopy (STEM) images can be a tedious task, if it has to be done manually. Therefore, a convolutional neural network is trained to do this task for STEM-images of TiO<sub>2</sub>-WO<sub>3</sub> nanoparticle hetero-aggregates. The present dataset contains the training data and some jupyter-notebooks that can be used after installation of the MMDetection toolbox (<a href="https://github.com/open-mmlab/mmdetection">https://github.com/open-mmlab/mmdetection</a>) to train the CNN. Details are provided in the manuscript submitted to Nano Select and in the comments of the jupyter-notebooks.</p> <p><strong>Authors and funding:</strong></p> <p>The present dataset was created by the authors. The work was funded by the Deutsche Forschungsgemeinschaft within the priority program SPP2289 under contract numbers RO2057/17-1 and MA3333/25-1.</p> <p><strong>Dataset description:</strong></p> <p>Four jupyter-notebooks are provided, which can be used for different tasks, according to their names. Details can be found within the comments and markdowns. These notebooks can be run after installation of MMDetection within the mmdetection folder.</p> <ul> <li><em>particle_detection_training.ipynb:</em> This notebook can be used for network training.</li> <li><em>particle_detection_evaluation.ipynb:</em> This notebook is for evaluation of a trained network with simulated test images.</li> <li><em>particle_detection_evaluation_experiment.ipynb:</em> This notebook is for evaluation of a trained network with experimental test images.</li> <li><em>particle_detection_measurement_experiment.ipynb:</em> This notebook is for application of a trained network to experimental data.</li> </ul> <p>In addition, a script titled <em>particle_detection_functions.py</em> is provided which contains functions required by the notebooks. Details can be found within the comments.</p> <p>The zip archive <em>training_data.zip</em> contains the training data. The subfolder <em>HAADF</em> contains the images (sorted as training, validation and test images), the subfolder <em>json </em>contains the annotation (sorted as training, validation and test images). Each file within the <em>json</em> folder provides for each image the following information:</p> <ul> <li>aggregat_no: image id, the number of the corresponding image file</li> <li>particle_position_x: list of particle position x-coordinates in nm</li> <li>particle_position_y: list of particle position y-coordinates in nm</li> <li>particle_position_z: list of particle position z-coordinates in nm</li> <li>particle_radius: list of volume equivalent particle radii in nm</li> <li>particle_type: list of material types, 1: TiO<sub>2</sub>, 2: WO<sub>3</sub></li> <li>particle_shape: list of particle shapes: 0: sphere, 1: box, 2: icosahedron</li> <li>rotation: list of particle rotations in rad. Each particle is rotated twice by the listed angle (before and after deformation)</li> <li>deformation: list of particle deformations. After the first rotation the particle x-coordinates of the particle’s surface mesh are scaled by the factor listed in deformation, y- and z-coordinates are scaled according to 1/sqrt(deformation).</li> <li>cluster_index: list of cluster indices for each particle</li> <li>initial_cluster_index: list of initial cluster indices for each particle, before primary clusters of the same material were merged</li> <li>fractal_dimension: the intended fractal dimension of the aggregate</li> <li>fractal_dimension_true: the realized geometric fractal dimension of the aggregate (neglecting particle densities)</li> <li>fractal_dimension_weight_true: the realized fractal dimension of the aggregate (including particle densities)</li> <li>fractal_prefactor: fractal prefactor</li> <li>mixing_ratio_intended: the intended mixing ratio (fraction of WO<sub>3</sub> particles)</li> <li>mixing_ratio_true: the realised mixing ratio (fraction of WO<sub>3</sub> particles)</li> <li>mixing_ratio_volume: the realised mixing ratio (fraction of WO<sub>3</sub> volume)</li> <li>mixing_ratio_weight: the realised mixing ratio (fraction of WO<sub>3</sub> weight)</li> <li>particle_1_rho: density of TiO<sub>2</sub> used for the calculations</li> <li>particle_1_size_mean: mean TiO<sub>2</sub> radius</li> <li>particle_1_size_min: smallest TiO<sub>2</sub> radius</li> <li>particle_1_size_max: largest TiO<sub>2</sub> radius</li> <li>particle_1_size_std: standard deviation of TiO<sub>2</sub> radii</li> <li>particle_1_clustersize: average TiO<sub>2</sub> cluster size</li> <li>particle_1_clustersize_init: average TiO<sub>2</sub> cluster size of primary clusters (before merging into larger clusters)</li> <li>particle_1_clustersize_init_intended: intended TiO<sub>2</sub> cluster size of primary clusters</li> <li>particle_2_rho: density of WO<sub>3 </sub>used for the calculations</li> <li>particle_2_size_mean: mean WO<sub>3</sub> radius</li> <li>particle_2_size_min: smallest WO<sub>3</sub> radius</li> <li>particle_2_size_max: largest WO<sub>3</sub> radius</li> <li>particle_2_size_std: standard deviation of WO<sub>3</sub> radii</li> <li>particle_2_clustersize: average WO<sub>3</sub> cluster size</li> <li>particle_2_clustersize_init: average WO<sub>3</sub> cluster size of primary clusters (before merging into larger clusters)</li> <li>particle_2_clustersize_init_intended: intended WO<sub>3</sub> cluster size of primary clusters</li> <li>number_of_primary_particles: number of particles within the aggregate</li> <li>gyration_radius_geometric: gyration radius of the aggregate (neglecting particle densities)</li> <li>gyration_radius_weighted: gyration radius of the aggregate (including particle densities)</li> <li>mean_coordination: mean total coordination number (particle contacts)</li> <li>mean_coordination_heterogen: mean heterogeneous coordination number (contacts with particles of the different material)</li> <li>mean_coordination_homogen: mean homogeneous coordination number (contacts with particles of the same material)</li> <li>radius_equiv: list of area equivalent particle radii (in projection)</li> <li>k_proj: projection direction of the aggregate: 0: z-direction (axis = 2), 1: x-direction (axis = 1), 2: y-direction (axis = 0)</li> <li>polygons: list of polygons that surround the particle (COCO annotation)</li> <li>bboxes: list of particle bounding boxes</li> <li>aggregate_size: projected area of the aggregate translated into the radius of a circle in nm</li> <li>n_pix: number of pixel per image in horizontal and vertical direction (squared images)</li> <li>pixel_size: pixel size in nm</li> <li>image_size: image size in nm</li> <li>add_poisson_noise: 1 if poisson noise was added, 0 otherwise</li> <li>frame_time: simulated frame time (required for poisson noise)</li> <li>dwell_time: dwell time per pixel (required for poisson noise)</li> <li>beam_current: beam current (required for poisson noise)</li> <li>electrons_per_pixel: number of electrons per pixel</li> <li>dose: electron dose in electrons per Å<sup>2</sup></li> <li>add_scan_noise: 1 if scan noise was added, 0 otherwise</li> <li>beam misposition: parameter that describes how far the beam can be misplaced in pm (required for scan noise)</li> <li>scan_noise: parameter that describes how far the beam can be misplaced in pixel (required for scan noise)</li> <li>add_focus_dependence: 1 if a focus effect is included, 0 otherwise</li> <li>data_format: data format of the images, e.g. uint8</li> </ul> <p>There are 24000 training images, 5500 validation images, 5500 test images, and their corresponding annotations. Aggregates and STEM images were obtained with the algorithm explained in the main work. The important data for CNN training is extracted from the files of individual aggregates and concluded in the subfolder <em>COCO</em>. For training, validation and test data there is a file <em>annotation_COCO.json</em> that includes all information required for the CNN training.</p> <p>The zip archive <em>experiment_test_data.zip</em> includes manually annotated experimental images. All experimental images were filtered as explained in the main work. The subfolder <em>HAADF</em> includes thirteen images. The subfolder <em>json</em> includes an annotation file for each image in COCO format. A single file concluding all annotations is stored in <em>json/COCO/annotation_COCO.json</em>.</p> <p>The zip archive <em>experiment_measurement.zip</em> includes the experimental images investigated in the manuscript. It contains four subfolders corresponding to the four investigated samples. All experimental images were filtered as explained in the manuscript.</p> <p>The zip archive <em>particle_detection.zip</em> includes the network, that was trained, evaluated and used for the investigation in the manuscript. The network weights are stored in the file <em>particle_detection/logs/fit/20230622-222721/iter_60000.pth</em>. These weights can be loaded with the jupyter-notebook files. Furthermore, a configuration file, which is required by the notebooks, is stored as <em>particle_detection/logs/fit/20230622-222721/config_file.py</em>.</p> <p>There is no confidential data in this dataset. It is neither offensive, nor insulting or threatening.</p> <p>The dataset was generated to discriminate between TiO<sub>2 </sub>and WO<sub>3</sub> nanoparticles in STEM-images. It might be possible that it can discriminate between different materials if the STEM contrast is similar to the contrast of TiO<sub>2 </sub>and WO<sub>3</sub> but there is no guarantee.</p>
Diagnostic Performance of a Convolutional Neural Network for Diminutive Colorectal Polyp Recognition
ClinicalTrials.gov study NCT03822390. IPD Sharing: UNDECIDED. Countries: 1. Publications: 2.
Data from: A convolutional neural network for detecting sea turtles in drone imagery
Open the record for dataset details and reuse information.
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.