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.
1,049
datasets available to search
ShareScore release 0.9.0
Dataset results
1,049 results for “Robustness”
Data for A Locally Activatable Sensor for Robust Quantification of Organellar Glutathione
<p>Supporting data to paper A Locally Activatable Sensor for Robust Quantification of Organellar Glutathione,</p> <p>including NMR, MS, microscopy, etc</p>
Data from: Skyline fossilized birth-death model is robust to violations of sampling assumptions in total-evidence dating
<p>Several total-evidence dating studies under the fossilized birth-death (FBD) model have produced very old age estimates, which are not supported by the fossil record. This phenomenon has been termed "deep root attraction (DRA)". For two specific datasets, involving divergence time estimation for the early radiations of ants, bees and wasps (Hymenoptera) and of placental mammals (Eutheria), it has been shown that the DRA effect can be greatly reduced by accommodating the fact that extant species in these trees have been sampled to maximize diversity, so called diversified sampling. Unfortunately, current methods to accommodate diversified sampling only consider the extreme case where it is possible to identify a cut-off time such that all splits occurring before this time are represented in the sampled tree but none of the younger splits. In reality, the sampling bias is rarely this extreme, and may be difficult to model properly. Similar modeling challenges apply to the sampling of the fossil record. This raises the question of whether it is possible to find dating methods that are more robust to sampling biases. Here, we show that the skyline FBD (SFBD) process, where the diversification and fossil-sampling rates can vary over time in a piecewise fashion, provides age estimates that are more robust to inadequacies in the modeling of the sampling process and less sensitive to DRA effects. In the SFBD model we consider, rates in different time intervals are either considered to be independent and identically distributed, or assumed to be autocorrelated following an Ornstein-Uhlenbeck (OU) process. Through simulations and reanalyses of the Hymenoptera and Eutheria data, we show that both variants of the SFBD model unify age estimates under random and diversified sampling assumptions. The SFBD model can resolve DRA by absorbing the deviations from the sampling assumptions into the inferred dynamics of the diversification process over time. Although this means that the inferred diversification dynamics must be interpreted with caution, taking sampling biases into account, we conclude that the SFBD model represents the most robust approach available currently for addressing DRA in total-evidence dating.</p>
Compound Data for Robust Processes for Polymer Modification and Pharmaceutical Synthesis
<p>Compound structural (IUPAC name, InChI, InChI Key, SMILES, .mol, .sdf) and spectral (NMR, MS) data included for compounds reported in the associated doctoral thesis. NMR data collected on Bruker Avance 400, 500, or 600 MHz spectrometers. Compound structure data were generated by ChemDraw v.20 (PerkinElmer). More details about the preparation and characterization of these compounds can be found in the associated thesis.</p>
Data for: A hypothesis for robust polarization vision: An example from the Australian Imperial Blue butterfly, Jalmenus evagoras
<p class="MsoNormal"><span>The Australian lycaenid butterfly, <em>Jalmenus evagoras</em>, has iridescent wings that are sexually dimorphic in both spectral reflection and degree of polarization, </span><span>suggesting</span><span> </span><span>that these wing properties are likely to be important in mate recognition. We first describe the results of a field experiment </span><span>showing<span> that free-flying individuals of <em>J. evagoras </em>discriminate between visual stimuli that vary in polarization content in blue wavelengths but not in others. We then present detailed reflectance spe</span>ctrophotometry</span><span> </span><span>measurements <span>of the polarization content of male and female wings, showing that female wings exhibit blue-shifted reflectance,</span> with</span><span> </span><span>a lower degree of polarization relative to male wings. </span><span>Finally, we describe a novel method for measuring alignment of ommatidial arrays:<span> </span>By measuring variation of depolarized eyeshine intensity from </span><span>patches of</span><span> ommatidia as a function of eye rotation, we show that </span><span>a) </span><span>individual </span><span>rhabdoms</span><span> contain mutually perpendicular microvilli</span><span>; b) many rhabdoms in the array</span><span> are rotated with respect to one another by as much as 45º</span><span>; c) the rotated ommatidia are useful for robust polarization detection. </span><span>By mapping the distribution of the ommatidial </span><span>rotations</span><span> in eye </span><span>patches</span><span> of <em>J. evagoras</em>, we show that males and females exhibit differences in the extent to which </span><span>ommatidia</span><span> are aligned</span><span>. Both</span><span> the </span><span>number</span><span> of </span><span>rotated </span><span>ommatidia suitable for </span><span>robust </span><span>polarization-detection</span><span>, and the number of aligned ommatidia suitable for</span><span> edge-detection</span><span>,</span><span> var</span><span>y</span><span> with respect to both sex and eye</span><span>-</span><span>patch elevation. </span><span>Thus, <em>J. evagoras </em>exhibits finely-tuned ommatidial arrays suitable for perception of polarized signals, likely to match sex-specific life history differences in the utility of polarized signals.</span></p>
ImageNet-Patch: A Dataset for Benchmarking Machine Learning Robustness against Adversarial Patches
<p>Adversarial patches are optimized contiguous pixel blocks in an input image that cause a machine-learning model to misclassify it. However, their optimization is computationally demanding and requires careful hyperparameter tuning. To overcome these issues, we propose ImageNet-Patch, a dataset to benchmark machine-learning models against adversarial patches. It consists of a set of patches optimized to generalize across different models and applied to ImageNet data after preprocessing them with affine transformations. This process enables an approximate yet faster robustness evaluation, leveraging the transferability of adversarial perturbations.</p> <p>We release our dataset as a set of folders indicating the patch target label (e.g., `banana`), each containing 1000 subfolders as the ImageNet output classes.</p> <p>An example showing how to use the dataset is shown below.</p> <pre><code class="language-python"># code for testing robustness of a model import os.path from torchvision import datasets, transforms, models import torch.utils.data class ImageFolderWithEmptyDirs(datasets.ImageFolder): """ This is required for handling empty folders from the ImageFolder Class. """ def find_classes(self, directory): classes = sorted(entry.name for entry in os.scandir(directory) if entry.is_dir()) if not classes: raise FileNotFoundError(f"Couldn't find any class folder in {directory}.") class_to_idx = {cls_name: i for i, cls_name in enumerate(classes) if len(os.listdir(os.path.join(directory, cls_name))) > 0} return classes, class_to_idx # extract and unzip the dataset, then write top folder here dataset_folder = 'data/ImageNet-Patch' available_labels = { 487: 'cellular telephone', 513: 'cornet', 546: 'electric guitar', 585: 'hair spray', 804: 'soap dispenser', 806: 'sock', 878: 'typewriter keyboard', 923: 'plate', 954: 'banana', 968: 'cup' } # select folder with specific target target_label = 954 dataset_folder = os.path.join(dataset_folder, str(target_label)) normalizer = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) transforms = transforms.Compose([ transforms.ToTensor(), normalizer ]) dataset = ImageFolderWithEmptyDirs(dataset_folder, transform=transforms) model = models.resnet50(pretrained=True) loader = torch.utils.data.DataLoader(dataset, shuffle=True, batch_size=5) model.eval() batches = 10 correct, attack_success, total = 0, 0, 0 for batch_idx, (images, labels) in enumerate(loader): if batch_idx == batches: break pred = model(images).argmax(dim=1) correct += (pred == labels).sum() attack_success += sum(pred == target_label) total += pred.shape[0] accuracy = correct / total attack_sr = attack_success / total print("Robust Accuracy: ", accuracy) print("Attack Success: ", attack_sr) </code></pre> <p> </p>
Lower complexity of motor primitives ensures robust control of high-speed human locomotion
<p>Walking and running are mechanically and energetically different locomotion modes. For selecting one or another, speed is a parameter of paramount importance. Yet, both are likely controlled by similar low-dimensional neuronal networks that reflect in patterned muscle activations called muscle synergies. Here, we investigated how humans synergistically activate muscles during locomotion at different submaximal and maximal speeds. We analysed the duration and complexity (or irregularity) over time of motor primitives, the temporal components of muscle synergies. We found that the challenge imposed by controlling high-speed locomotion forces the central nervous system to produce muscle activation patterns that are wider and less complex relative to the duration of the gait cycle. The motor modules, or time-independent coefficients, were redistributed as locomotion speed changed. These outcomes show that robust locomotion control at challenging speeds is achieved by modulating the relative contribution of muscle activations and producing less complex and wider control signals, whereas slow speeds allow for more irregular control.</p> <p> </p> <p>In this supplementary data set we made available: a) the metadata with anonymized participant information, b) the raw EMG, c) the touchdown and lift-off timings of the recorded limb, d) the filtered and time-normalized EMG, e) the muscle synergies extracted via NMF and f) the code to process the data, including the scripts to calculate the Higuchi's fractal dimension (HFD) of motor primitives. In total, 180 trials from 30 participants are included in the supplementary data set.</p> <p>The file “metadata.dat” is available in ASCII and RData format and contains:</p> <ul> <li>Code: the participant’s code</li> <li>Group: the experimental group in which the participant was involved (G1 = walking and submaximal running; G2 = submaximal and maximal running)</li> <li>Sex: the participant’s sex (M or F)</li> <li>Speeds: the type of locomotion (W for walking or R for running) and speed at which the recordings were conducted in 10*[m/s]</li> <li>Age: the participant’s age in years</li> <li>Height: the participant’s height in [cm]</li> <li>Mass: the participant’s body mass in [kg]</li> <li>PB: 100 m-personal best time (for G2).</li> </ul> <p>The "RAW_DATA.RData" R list consists of elements of S3 class "EMG", each of which is a human locomotion trial containing cycle segmentation timings and raw electromyographic (EMG) data from 13 muscles of the right-side leg. Cycle times are structured as data frames containing two columns that correspond to touchdown (first column) and lift-off (second column). Raw EMG data sets are also structured as data frames with one row for each recorded data point and 14 columns. The first column contains the incremental time in seconds. The remaining 13 columns contain the raw EMG data, named with the following muscle abbreviations: ME = gluteus medius, MA = gluteus maximus, FL = tensor fasciæ latæ, RF = rectus femoris, VM = vastus medialis, VL = vastus lateralis, ST = semitendinosus, BF = biceps femoris, TA = tibialis anterior, PL = peroneus longus, GM = gastrocnemius medialis, GL = gastrocnemius lateralis, SO = soleus. Please note that the following trials include less than 30 gait cycles (the actual number shown between parentheses): P16_R_83 (20), P16_R_95 (25), P17_R_28 (28), P17_R_83 (24), P17_R_95 (13), P18_R_95 (23), P19_R_95 (18), P20_R_28 (25), P20_R_42 (27), P20_R_95 (25), P22_R_28 (23), P23_R_28(29), P24_R_28 (28), P24_R_42 (29), P25_R_28 (29), P25_R_95 (28), P26_R_28 (29), P26_R_95 (28), P27_R_28 (28), P27_R_42 (29), P27_R_95 (24), P28_R_28 (29), P29_R_95 (17). All the other trials consist of 30 gait cycles. Trials are named like “P20_R_20,” where the characters “P20” indicate the participant number (in this example the 20th), the character “R” indicate the locomotion type (W=walking, R=running), and the numbers “20” indicate the locomotion speed in 10*m/s (in this case the speed is 2.0 m/s). The filtered and time-normalized emg data is named, following the same rules, like “FILT_EMG_P03_R_30”.</p> <p><strong>Old versions not compatible with the R package <a href="https://CRAN.R-project.org/package=musclesyneRgies">musclesyneRgies</a></strong></p> <p>The files containing the gait cycle breakdown are available in RData format, in the file named “CYCLE_TIMES.RData”. The files are structured as data frames with as many rows as the available number of gait cycles and two columns. The first column named “touchdown” contains the touchdown incremental times in seconds. The second column named “stance” contains the duration of each stance phase of the right foot in seconds. Each trial is saved as an element of a single R list. Trials are named like “CYCLE_TIMES_P20_R_20,” where the characters “CYCLE_TIMES” indicate that the trial contains the gait cycle breakdown times, the characters “P20” indicate the participant number (in this example the 20th), the character “R” indicate the locomotion type (W=walking, R=running), and the numbers “20” indicate the locomotion speed in 10*m/s (in this case the speed is 2.0 m/s). Please note that the following trials include less than 30 gait cycles (the actual number shown between parentheses): P16_R_83 (20), P16_R_95 (25), P17_R_28 (28), P17_R_83 (24), P17_R_95 (13), P18_R_95 (23), P19_R_95 (18), P20_R_28 (25), P20_R_42 (27), P20_R_95 (25), P22_R_28 (23), P23_R_28(29), P24_R_28 (28), P24_R_42 (29), P25_R_28 (29), P25_R_95 (28), P26_R_28 (29), P26_R_95 (28), P27_R_28 (28), P27_R_42 (29), P27_R_95 (24), P28_R_28 (29), P29_R_95 (17).</p> <p>The files containing the raw, filtered and the normalized EMG data are available in RData format, in the files named “RAW_EMG.RData” and “FILT_EMG.RData”. The raw EMG files are structured as data frames with as many rows as the amount of recorded data points and 13 columns. The first column named “time” contains the incremental time in seconds. The remaining 12 columns contain the raw EMG data, named with muscle abbreviations that follow those reported above. Each trial is saved as an element of a single R list. Trials are named like “RAW_EMG_P03_R_30”, where the characters “RAW_EMG” indicate that the trial contains raw emg data, the characters “P03” indicate the participant number (in this example the 3rd), the character “R” indicate the locomotion type (see above), and the numbers “30” indicate the locomotion speed (see above). The filtered and time-normalized emg data is named, following the same rules, like “FILT_EMG_P03_R_30”.</p> <p>The files containing the muscle synergies extracted from the filtered and normalized EMG data are available in RData format, in the files named “SYNS_H.RData” and “SYNS_W.RData”. The muscle synergies files are divided in motor primitives and motor modules and are presented as direct output of the factorisation and not in any functional order. Motor primitives are data frames with 6000 rows and a number of columns equal to the number of synergies (which might differ from trial to trial) plus one. The rows contain the time-dependent coefficients (motor primitives), one column for each synergy plus the time points (columns are named e.g. “time, Syn1, Syn2, Syn3”, where “Syn” is the abbreviation for “synergy”). Each gait cycle contains 200 data points, 100 for the stance and 100 for the swing phase which, multiplied by the 30 recorded cycles, result in 6000 data points distributed in as many rows. This output is transposed as compared to the one discussed in the methods section to improve user readability. Each set of motor primitives is saved as an element of a single R list. Trials are named like “SYNS_H_P12_W_07”, where the characters “SYNS_H” indicate that the trial contains motor primitive data, the characters “P12” indicate the participant number (in this example the 12th), the character “W” indicate the locomotion type (see above), and the numbers “07” indicate the speed (see above). Motor modules are data frames with 12 rows (number of recorded muscles) and a number of columns equal to the number of synergies (which might differ from trial to trial). The rows, named with muscle abbreviations that follow those reported above, contain the time-independent coefficients (motor modules), one for each synergy and for each muscle. Each set of motor modules relative to one synergy is saved as an element of a single R list. Trials are named like “SYNS_W_P22_R_20”, where the characters “SYNS_W” indicate that the trial contains motor module data, the characters “P22” indicate the participant number (in this example the 22nd), the character “W” indicates the locomotion type (see above), and the numbers “20” indicate the speed (see above). Given the nature of the NMF algorithm for the extraction of muscle synergies, the supplementary data set might show non-significant differences as compared to the one used for obtaining the results of this paper.</p> <p>The files containing the HFD calculated from motor primitives are available in RData format, in the file named “HFD.RData”. HFD results are presented in a list of lists containing, for each trial, 1) the HFD, and 2) the interval time <em>k</em> used for the calculations. HFDs are presented as one number (mean HFD of the primitives for that trial), as are the interval times <em>k</em>. Trials are named like “HFD_P01_R_95”, where the characters “HFD” indicate that the trial contains HFD data, the characters “P01” indicate the participant number (in this example the 1st), the character “R” indicates the locomotion type (see above), and the numbers “95” indicate the speed (see above).</p> <p>All the code used for the pre-processing of EMG data, the extraction of muscle synergies and the calculation of HFD is available in R format. Explanatory comments are profusely present throughout the script “muscle_synergies.R”.</p>
Data from: Robust Estimation of Field Inhomogeneity Map Following Magnitude-Based Water-Fat Separation with Resolved Ambiguity
<p>These data have been uploaded and shared as part of "Robust Estimation of Field Inhomogeneity Map Following Magnitude-Based Water-Fat Separation with Resolved Ambiguity". The data may be used for field inhomogeneity mapping and PDFF/R2* reconstruction. </p> <p>These data were acquired by Perspectum Ltd (https://perspectum.com/) on a healthy volunteer. Informed consent was obtained from the participant. The dataset includes a localizer series and a multi-slice series (magnitude and phase) acquired from a volunteer covering the dome of the liver, heart and lungs:</p> <ul> <li>1-localizer_haste_bh</li> <li>2-I_6_Echo_3D_32_Slice_IDEAL</li> <li>3-I_6_Echo_3D_32_Slice_IDEAL</li> </ul> <p>These data were gathered using a Siemens Prisma 3 Tesla scanner. The main dataset comprises an acquisition with thirty-two slices including the abdominal region, with slices placed away from the isocenter. The acquisition consisted of a 6‐echo (TE1=1.3 ms, ΔTE=1 ms) gradient-recalled echo (GRE) protocol designed to minimize T1 bias (3° flip angle), Pixel Bandwidth = 1565 Hz, and 232 x 256 reconstructed image size, with 5 mm slice thickness and 1.72 x 1.72 mm^2 in-plane resolution. </p>
Data set for the article 'Robust replication initiation from coupled homeostatic mechanisms'
<p>This data set contains the data of the submitted article "Robust replication initiation from coupled homeostatic mechanisms". The data was generated using simulations in python that are linked below. Experiments indicate that E. coli controls replication initiation via titration and activation of the initiator protein DnaA. We study by mathematical modelling how these two mechanisms interact to generate robust replication-initiation cycles. </p>
Model Zoo for Robust Models are less Over-Confident
<p><strong>Model Zoo (PyTorch) of non-adversarially trained models for Robust Models are less Over-Confident (NeurIPS'22)</strong></p> <p>Abstract: <em>"Regardless of the success of convolutional neural networks (CNNs) in many academic benchmarks of computer vision tasks, their application in real-world is still facing fundamental challenges, like the inherent lack of robustness as unveiled by adversarial attacks. These attacks target to manipulate the network's prediction by adding a small amount of noise onto the input. In turn, adversarial training (AT) aims to achieve robustness against such attacks by including adversarial samples in the trainingset. However, a general analysis of the reliability and model calibration of these robust models beyond adversarial robustness is still pending. In this paper, we analyze a variety of adversarially trained models that achieve high robust accuracies when facing state-of-the-art attacks and we show that AT has an interesting side-effect: it leads to models that are significantly less overconfident with their decisions even on clean data than non-robust models. Further, our analysis of robust models shows that not only AT but also the model's building blocks (activation functions and pooling) have a strong influence on the models' confidence."</em></p>
Datasets-A globally robust relationship between water table decline, subsidence rate and carbon release from peatlands
<p>Supplementary Data A shows meta-data for in-situ and laboratory measurements of soil respiration or its components soil heterotrophic respiration and autotrophic respiration, as well as associated environmental variables from global pristine peatlands and water table decline peatlands, respectively.</p> <p> </p> <p>Supplementary Data B shows the relationships between peatland subsidence rates and drainage years for different land uses in different climate zones, relationships between proportion of peatland subsidence rates due to oxidation and drainage years for different land uses in different climate zones, the estimated peat subsidence rates and peat subsidence rates due to oxidation, the synthesized soil organic carbon content and soil bulk density at the layer of 0-30 cm from pristine peatlands, and the in-situ measured annual soil heterotrophic respiration rates for validating the robustness of the developed emipirical models of this study. </p>
Data from: Treated like dirt: Robust forensic and ecological inferences from soil eDNA after challenging sample storage
<p>We investigated the effect of storage duration and conditions on the assessment of the soil biota with eDNA metabarcoding. We extracted eDNA from freshly collected soil samples and again from the same samples after storage under contrasting temperature conditions and contrasting exposure (open/closed tubes). We used four different primer sets targeting bacteria, fungi, protists (cercozoans), and general eukaryotes. <span>W</span>e quantified differences in richness, evenness, and community composition. Subsequently, we tested whether we could correctly infer habitat type and original sample identity after storage using a large reference dataset.</p> <p>This repository contains the un-demultiplexed fastq sequences.</p>
SCM-CNN: A Robust Deep Learning Matting Model for Cloud Removal in Optical Imagery
<p>This is a dataset that can be used for cloud detection and cloud opacity estimation. The data is saved in python-numpy form, and stored in dictionary :dict_keys(['OriginImage', 'Gimage', 'Alpha', 'Trimap', 'CloudMaxDN']) represents the cloud-free remote sensing image, cloud remote sensing image, cloud opacity, trilateration information and cloud brightness respectively. The command {np.load("Path",allow_pickle=True).item()} is used to read, where "Path" is the corresponding path to the file.</p>
Raw images and processed datasets related to the journal article Robust Assessment of Post-Localisation Hardening Behaviour in Eurofer97 using Inverse Finite Element Methods
Open the record for dataset details and reuse information.
Robust retrieval of forest canopy structural attributes using multi-platform airborne LiDAR
<p><strong>Data and R code to replicate the analyses presented in</strong>:<br>Zhang et al. (2024) Robust retrieval of forest canopy structural attributes using multi-platform airborne LiDAR. Remote Sensing in Ecology and Conservation, <a href="https://doi.org/10.1002/rse2.398">https://doi.org/10.1002/rse2.398</a></p> <p>If using these data and/or R code in your work please cite the original publication listed above, as well as this repository using the corresponding DOI.</p>
Robust quantum dots charge autotuning using neural network uncertainty - Output data
<p>Outputs of the model training and the offline autotuning experiments presented in the paper: "<em>Robust quantum dots charge autotuning using neural network uncertainty</em>".</p> <p>For convenience, the results are splitted in several zipped files:</p> <ul> <li><strong>run_outputs_light.zip</strong>: contains only settings and results text files (sufficient for compiling result tables).</li> <li><strong>run_outputs_full_scan.zip</strong>: contains complete scan of the diagrams (for qualitative analyse)</li> <li><strong>run_outputs_part<N>.zip</strong>: contains all autotuning simulation output, grouped by seed (images and video output types might vary between seeds)</li> </ul> <p>Each folder in the zipped files represent a run that includes:</p> <ul> <li>log file</li> <li>plots / images</li> <li>run settings</li> <li>performance results</li> <li>pytorch model parameters</li> </ul> <p>See README.txt for more information about the file strucutre.</p>
Data for Robust Algorithms for the analysis of the FFC-NMRD dispersion curves
<p>Data for article </p>
Illumina Sequencing Data for "Elucidating human gut microbiota interactions that robustly inhibit diverse Clostridioides difficile strains across different nutrient landscapes"
<p>Illumina Sequencing Data for Sulaiman et al., "Elucidating human gut microbiota interactions that robustly inhibit diverse Clostridioides difficile strains across different nutrient landscapes".</p>
Fig. 1 in Testing the robustness of transmission network models to predict ectoparasite loads. One lizard, two ticks and four years
Fig. 1. Transmission networks generated with (a) a short time window of infection; and (b) a long time window of infection, from the GPS location data of the lizards in the study population in 2010. Nodes represent individual lizards and edges between nodes are directed towards the lizard that is at risk of infection. The edges are weighted as described in the main text and the thicker the line the more weight is associated with that edge.
MedMNIST-C: Comprehensive benchmark and improved classifier robustness by simulating realistic image corruptions
<p><strong>Abstract: </strong>The integration of neural-network-based systems into clinical practice is limited by challenges related to domain generalization and robustness. The computer vision community established benchmarks such as ImageNet-C as a fundamental prerequisite to measure progress towards those challenges. Similar datasets are largely absent in the medical imaging community which lacks a comprehensive benchmark that spans across imaging modalities and applications. To address this gap, we create and open-source MedMNIST-C, a benchmark dataset based on the MedMNIST+ collection, covering 12 datasets and 9 imaging modalities. We simulate task and modality-specific image corruptions of varying severity to comprehensively evaluate the robustness of established algorithms against real-world artifacts and distribution shifts. We further provide quantitative evidence that our simple-to-use artificial corruptions allow for highly performant, lightweight data augmentation to enhance model robustness. Unlike traditional, generic augmentation strategies, our approach leverages domain knowledge, exhibiting significantly higher robustness when compared to widely adopted methods. By introducing MedMNIST-C and open-sourcing the corresponding library allowing for targeted data augmentations, we contribute to the development of increasingly robust methods tailored to the challenges of medical imaging. The code is available at <a href="https://github.com/francescodisalvo05/medmnistc-api">github.com/francescodisalvo05/medmnistc-api</a>.</p> <blockquote> <p>This work has been accepted at the Workshop on Advancing Data Solutions in Medical Imaging AI @ MICCAI 2024 [<a href="https://arxiv.org/pdf/2406.17536" target="_blank" rel="noopener">preprint</a>].</p> </blockquote> <p><strong>Note: </strong>Due to space constraints, we have uploaded all datasets except TissueMNIST-C. However, it can be reproduced via our APIs. </p> <p><strong>Usage: </strong>We recommend using the demo code and tutorials available on our GitHub <a href="https://github.com/francescodisalvo05/medmnistc-api">repository</a>.</p> <p><strong>Citation: </strong>If you find this work useful, please consider citing us:</p> <blockquote> <pre>@article{disalvo2024medmnist, title={MedMNIST-C: Comprehensive benchmark and improved classifier robustness by simulating realistic image corruptions}, author={Di Salvo, Francesco and Doerrich, Sebastian and Ledig, Christian}, journal={arXiv preprint arXiv:2406.17536}, year={2024} }</pre> </blockquote> <p><strong>Disclaimer: </strong>This repository is inspired by MedMNIST APIs and the ImageNet-C repository. Thus, please also consider citing <a href="https://www.nature.com/articles/s41597-022-01721-8" rel="nofollow">MedMNIST</a>, the respective source datasets (described <a href="https://medmnist.com/" rel="nofollow">here</a>), and <a href="https://arxiv.org/abs/1903.12261" rel="nofollow">ImageNet-C</a>.</p>
Figure 6. (a1), (a2), (a3), (a4), (a5), (a6), (a7) and (a8) watermarked image is degraded respectively through JPEG2000 compression, JPEG compression, median filtering, adding Salt&Pepper noise, rotating, center cropping, surrounding cropping and scaling. (b1), (b2), (b3), (b4), (b5), (b6), (b7) and (b8) The corresponding extracted watermarks.-Discrete Wavelet Transform Method: A New Optimized Robust Digital Image Watermarking Scheme
<p>This paper has described a scheme for digital watermarking of still images based on discrete<br> wavelet transform. In the proposed method, the embedded logo watermark can be extracted without<br> access to the original image. It has been confirmed that the proposed watermarking method is able<br> to extract the embedded logo watermark from the watermarked images that have degraded through<br> compression, filtering, cropping and scaling. Although this algorithm is not robust against rotation,<br> it can completely extract the watermark from watermarked images that lose about 35% of their<br> areas by cropping attack.</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.