Find research datasets worth reusing

opencc-zeroMar 2023View details →

Data from: Taking eDNA underground: factors affecting eDNA detection of subterranean fauna in groundwater

Stygofauna are ancient aquatic fauna that have evolved to live underground. The impacts of anthropogenic climate change, extraction and pollution on groundwater pose major threats to groundwater health, prompting the need for efficient and reliable means to detect and monitor stygofaunal communities. Traditional survey techniques for these species rely on morphological identification and can be biased, labour intensive, and often indeterminate to lower taxonomic levels. By contrast, environmental DNA (eDNA)-based methods have the potential to dramatically improve on existing stygofaunal survey methods in a large range of habitats and for all life stages, reducing the need for the destructive manual collection of often critically endangered species or specialized taxonomic expertise. We compared eDNA and haul-net samples collected in 2020 and 2021 from 19 groundwater bores and a cave on Barrow Island, located northwest of Western Australia, and assessed how sampling factors influenced the quality of eDNA detection of stygofauna. The two detection methods were complementary, eDNA metabarcoding was able to detect soft-bodied taxa and fish often missed by nets, but only detected seven of the nine stygofaunal crustacean orders identified from haul-net specimens. Our results also indicated that eDNA metabarcoding could detect 54-100% of stygofauna from shallow water samples and 82-90% from sediment samples. However, there was significant variation in stygofauna diversity between sample years and sampling types. The findings of this study demonstrate that haul net sampling has a tendency to underestimate stygofaunal diversity and that eDNA metabarcoding of groundwater can substantially improve efficiency of stygofaunal surveys.

opencc-by-4.0Mar 2023View details →

Detection of slow-moving landslides from Sentinel-2 optical satellite imagery

Datasets and code associated with recent ESPL publication "Detection of slow-moving landslides through automated satellite monitoring of surface deformation" The three imagery folders are zipped for convenience. The full code is included in the one .m file. Please get in touch with any questions.

opencc-zeroApr 2023View details →

Data from: Detection of vertebrates from natural and artificial inland water bodies in a semi-arid habitat using eDNA from filtered, swept and sediment samples

Climate warming will impact the sustainability of arid and semi-arid zone environments so we need to understand the influence of changes in arid lands on vertebrate populations. However, biomonitoring and biodiversity assessment in arid environments can be prohibitively time-consuming, expensive, and logistically challenging due to their often remote and inhospitable nature. Sampling of environmental DNA (eDNA) coupled with high-throughput sequencing is an emerging biodiversity assessment method. Here we explore the application of eDNA metabarcoding and various sampling approaches to estimate vertebrate richness and assemblage at human-constructed and natural water sources in a semi-arid region of Western Australia. Three sampling methods: sediment samples, filtering through a membrane with a pump, and membrane sweeping in the water body, were compared using two eDNA metabarcoding assays, 12S-V5 and 16smam, for 120 eDNA samples collected from four gnammas (gnamma: Australian Indigenous Noongar language term – granite rock pools) and four cattle troughs in the Great Western Woodlands, Western Australia. We detected higher vertebrate richness in samples from cattle troughs and found differences between assemblages detected in gnammas (more birds and amphibians) and cattle troughs (more mammals, including feral taxa). Total vertebrate richness was not different between swept and filtered samples, but all sampling methods yielded different assemblages. Our findings indicate that eDNA surveys in arid lands will benefit from collecting multiple samples at multiple water sources to avoid underestimating vertebrate richness. The high concentration of eDNA in small, isolated water bodies permits the use of sweep sampling which simplifies sample collection, processing, and storage, particularly when assessing vertebrate biodiversity across large spatial scales.

opencc-by-4.0Apr 2023View details →

Multi-topography dataset for wind turbine detection from remote sensing image

The land remote sensing wind turbine dataset has 1270 remote sensing images and contains 4459 individual wind turbines.  The images are taken over a large time span and contain remote sensing images of the same wind farm at different times. The dataset has both YOLO and VOC tagging formats. The dataset can be divided into five categories based on the land background: sandy land, forest land, grassland, snow land, and wasteland. Rich land background can improve the robustness of the model, and different marker formats and a large number of wind turbine individuals can meet the object detection of different models. Affected by the size of different power wind turbines, the multi-angle imaging characteristics of remote sensing satellites, the different solar radiation angles in different seasons and vegetation shading, wind turbines show large differences in the images. The image features of the wind turbine shadow are more obvious than those of the wind turbine body, so in order for the detection model to better identify the wind turbine, we label the wind turbine body and the wind turbine shadow as a whole when using the labelImg tool for labeling the wind turbine target.

opencc-zeroApr 2023View details →

Detecting ecological traps in human-altered landscapes: A case study of the thick-billed longspur nesting in croplands

Conversion of the North American prairies to cropland remains a prominent threat to grassland bird populations. Yet, a few species nest in these vastly modified systems. The thick-billed longspur (Rhynchophanes mccownii) is an obligate grassland bird whose populations have declined 4% annually during the past 50 years. Thick-billed longspurs historically nested in recently disturbed or sparsely vegetated patches within native mixed-grass prairie, but observations of longspurs in spring cereal and pulse crop fields during the breeding season in northeastern Montana, USA suggest such fields also provide cues for habitat selection. Maladaptive selection for poor-quality habitat may contribute to ongoing declines in longspur populations, but information on thick-billed longspur breeding ecology in crop fields is lacking. We hypothesized that these crop fields may function as ecological traps; specifically, we expected that crop fields may provide cues for territory selection, but frequent human disturbance and increased exposure to weather and predators would have negative consequences for reproduction. To address this hypothesis, we compared measures of habitat selection (settlement patterns and trends in abundance) and productivity (nest density, nest survival, and number of young fledged) between crop fields and native grassland sites during 2020–21. Across both years, settlement patterns were similar between site types and occupancy ranged from 0.52 ± 0.17SE to 0.99 ± 0.01 on April 7 and 30, respectively. Early season abundance differed by year, and changes in abundance during the breeding season appeared to be associated with precipitation-driven vegetation conditions rather than habitat type. While an index of nest density was lower in crop than native sites, the number of young fledged per successful nest (2.9 ± 0.18SE) and nest survival (0.24 ± 0.03 SE; n=222 nests) were similar for crop and native sites. Collectively, the data did not support our ecological trap hypothesis: longspurs did not exhibit a clear preference for crop sites and reproductive output was not significantly reduced. Our results indicate that croplands may provide alternative breeding habitat within a human-dominated landscape.

opencc-byApr 2023View details →

First Results on Wake Detection in SAR Images by Deep Learning

Dataset compendium from <a href="https://doi.org/10.3390/rs13224573">https://doi.org/10.3390/rs13224573</a>.

opencc-by-4.0Apr 2023View details →

Adapting camera-trap placement based on animal behaviour for rapid detection: a focus on the Endangered, white-bellied pangolin (Phataginus tricuspis)

Table containing detection data of species using two camera trap placement strategies (log vs non-log)

opencc-by-4.0Apr 2023View details →

DCASE 2023 Task5: Few-shot Bioacoustic Event Detection: Evaluation set

General Description The evaluation set for task 5 of DCASE 2023 "Few-shot Bioacoustic Event Detection" consists of 8 subsets of data representing different acoustic sources, in total there are 66 audio files.  The first 5 annotations are provided for each file, with events marked as positive (POS) for the class of interest.  This dataset is to be used for evaluation purposes during the task.   Folder structure Audiofiles and annotation files are organized across 3 different zip files: Eval_1.zip     |___CHE23/         |____*.wav         |____*.csv     |___CW/         |____*.wav         |____*.csv Eval_2.zip     |___MGE/         |____*.wav         |____*.csv     |___MS/         |____*.wav         |____*.csv     |___QU/         |____*.wav         |____*.csv    Eval_3.zip     |___DC/         |____*.wav         |____*.csv     |___CT/         |____*.wav         |____*.csv     |___CHE/         |____*.wav         |____*.csv Annotations_only.zip contains the *.csv files organised in the same subfolders. *The subfolders denote different recording sources and there may or may not be overlap between classes of interest from different wav files. *Note that there can be different target classes within the same subfolder   Annotation structure Each line of the annotation csv represents an event in the audio file. The column descriptions are as follows: [ Audiofilename, Starttime, Endtime, Q ] Development Set The development set for the same task can be found at: <a href="http://doi.org/10.5281/zenodo.4543504">https://doi.org/10.5281/zenodo.6012309</a>.  Open Access This dataset is available under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.   Contact info Please send any feedback or questions to: Ines Nolasco: i.dealmeidanolasco@qmul.ac.uk or join us on slack: <a href="https://join.slack.com/t/dcase/shared_invite/zt-12zfa5kw0-dD41gVaPU3EZTCAw1mHTCA">task-fewshot-bio-sed</a>

opencc-zeroMay 2023View details →

Experimental feeding validates nanofluidic array technology for DNA detection of ungulate prey in wolf scats

The study of carnivores' diet is a key component to enhance knowledge on the ecology of predators and their effect on prey populations. Although molecular approaches to detect prey DNA in carnivore scats are improving, the validation of their accuracy, a prerequisite for reliable applications within ecological frameworks, is still lagging behind the methodological advances. Indeed, variation in detection probability among prey species can occur, representing a potentially insidious source of bias in food-habit studies of carnivores. Calibration of DNA-based methods involves the optimization of specificity and sensitivity and, whereas priority is usually given to the former to avoid false positives, sensitivity is rarely investigated so that false negatives may be largely overlooked. We conducted feeding trials with captive wolves (Canis lupus) to validate a nanofluidic array technology recently developed for detection of multiple prey species in scats. Using 371 scat samples from 12 wolves fed with a single-prey diet, the sensitivity of our nanofluidic array method varied between 0.45 and 0.95 for the six main ungulate prey species. The method sensitivity was enhanced by using multiple markers per species and by a relatively low threshold for the number of amplifying markers required to confirm a detection. Yet, at least two markers should be used to avoid false positives. By acknowledging sources of bias in sensitivity to reliably interpret results of DNA-based dietary methods, our study highlights the relevance of feeding experiments to optimally calibrate the relative thresholds to define a positive detection and investigate occurrence and extent of biases in sensitivity.

Example testing images for PFU detection neural network

This is the example testing images used for the paper of "Rapid and stain-free quantification of viral plaque via lens-free holography and deep learning". The data has been compressed into eight zip files from .001 to .008. To access the data, please follow these steps: 1) Download all eight zipped files. 2) Once the download is complete, click on any one of the zipped files to start the extraction process. 3) The extracted files will automatically be organized into one folder. After extraction, there are three subfolders inside it: 1) Network input: it contained 4 holographic phase images (.mat files) at 12h, 13h, 14h and 15h of incubation for an example postive well and an example negative well. 2) Network output: It contains the network output image (PFU probability map) fot the example postive well and the example negative well. 3) Detection result: It contains the final binary detection result images after thresholding fot the example postive well and the example negative well.

opencc-by-4.0Dec 2022View details →

Mode characterization and sensitivity evaluation of an ultra-high-frequency surface acoustic wave (UHF-SAW) resonator biosensor: application to the glial-fibrillary-acidic-protein (GFAP) biomarker detection

Biosensors detect specific bio-analytes by generating a measurable signal from the interaction between the sensing element and the target molecule. Surface acoustic wave (SAW) biosensors offer unique advantages due to their high sensitivity, real-time response capability, and label-free detection. The typical SAW modes are the Rayleigh mode and the shear-horizontal mode. Both present pros and cons for biosensing applications and generally need different substrates and device geometries to be efficiently generated. This study investigates and characterizes ultra-high-frequency (UHF-) SAW resonator biosensors. It reveals the simultaneous presence of the two typical SAW modes, clearly separated in frequency, called slow and fast. The two modes are studied by numerical simulations and biosensing experiments with the glial-fibrillary-acidic-protein (GFAP) biomarker. The slow mode is generally more sensitive to changes in surface properties, such as temperature and mass changes, by a factor of about 1.4 with respect to the fast mode.

Modified Sen1Floods11 Dataset for Change Detection

This is the dataset used for flood mapping as a change detection task in the IGARSS 22 conference paper "Attentive Dual Stream Siamese U-Net for Flood Detection on Multi-Temporal Sentinel-1 Data" Paper Abstract: Due to climate and land-use change, natural disasters such as flooding have been increasing in recent years. Timely and reliable flood detection and mapping can help emergency response and disaster management. In this work, we propose a flood detection network using bi-temporal SAR acquisitions. The proposed segmentation network has an encoder-decoder architecture with two Siamese encoders for pre and post-flood images. The network's feature maps are fused and enhanced using attention blocks to achieve more accurate detection of the flooded areas. Our proposed network is evaluated on publicly available Sen1Flood11 [1] benchmark dataset. The network outperformed the existing state-of-the-art (uni-temporal) flood detection method by 6% IOU. The experiments highlight that the combination of bi-temporal SAR data with an effective network architecture achieves more accurate flood detection than uni-temporal methods.  Full Paper : <a href="https://ieeexplore.ieee.org/document/9883132">https://ieeexplore.ieee.org/document/9883132</a>

Environmental DNA detection range for Hydrilla

Dataset associated with publication of the same name including qPCR results, measured variables, and other supplemental data.

Dataset of manuscript "No detectable upper limit of mineral associated carbon in temperate agricultural soils"

Dataset of carbon fractions (mineral associated organic carbon and particulate organic carbon) detected in a subset of topsoil samples of the first German Agricultural Soil Inventory.

opencc-zeroMay 2023View details →

Data from: Hiding among colors: Background color diversity impedes detection time

Avoiding detection is vital for the survival of many animals. Factors extrinsic to animals, such as the visual complexity of the background, have been shown to impede the detection of animals. Studies using artificial and natural backgrounds have attributed background complexity to various visual features of the background. One feature that has received less attention is the diversity of color (hue) in the background. We used chickens and artificial backgrounds containing perceptually distinct elements in experiments to test whether color and luminance diversity affect detection time. We found that color diversity in the background impeded detection, while color diversity in prey and luminance diversity in the background did not impede detection. We also did not find an effect of luminance contrast on detection time. Our study suggests a prey animal can benefit in terms of increased detection times by predators when resting on backgrounds with enhanced color diversity.

Out-of-distribution detection algorithms for robust insect classification dataset and models

This folder contains trained models and datasets for reproducing the results in the paper on out-of-distribution detection algorithms for robust insect classification. Specifically, it contains the following folders:    <ul> <li>OODInsect (out-of-distribution data)</li> <li>MSP, MAH, and EBM trained models, each wrapped around the three classifiers of ResNet50, RegNet32, and VGG11, and different combinations of ID and OOD test data for reproducing RQ1, RQ2, and RQ3.</li> <li>ID3 (in-distribution test data)</li> </ul>

Supplemental data for: Visualization of rank-citation curves for fast detection of possible manipulations with the h-index of the university

This dataset consists of papers of universities in the top 30 Scopus Ranking of Ukrainian Universities (May 2023). The data was obtained from Scopus using the search query "AF-ID (“university name”) AND PUBYEAR < 2023 AND PUBYEAR > 2002". Rank-citation curves were also generated for the publications of each university. In this analysis, the rank of publications was plotted along the horizontal axis, while the corresponding citation counts were depicted on the left axis. All types of documents were included in the dataset.

Validation of a SARS-CoV-2 surrogate neutralization test detecting neutralizing antibodies against the major variants of concern

SARS-CoV-2 infection and/or vaccination elicit a broad range of neutralizing antibody responses against the different variants of concern (VOC). We established a new variant-adapted surrogate neutralization test (sVNT) and assessed the neutralization activity against the ancestral B.1 (WT) as well as VOC Delta, Omicron BA.1, BA.2 and BA.5. Performances were compared to the reference virus neutralization test (VNT) against the respective VOC using three different cohorts collected during the COVID waves. Pre-COVID samples confirmed 100% specificity of the sVNT. Correlation analyses showed moderate to strong correlation for Omicron sub-variants (Spearman’s r=0.7081 for BA.1, r=0.7205 for BA.2 and r=0.6042 for BA.5), and for WT (r=0.8458) and Delta-sVNT (r=0.8158), respectively. Comparison of the WT-sVNT performance with two CE-IVD commercial kits “Icosagen SARS-CoV-2 Neutralizing Antibody ELISA kit” and “Genscript cPass kit” revealed an overall good correlation ranging from 0.8673 to -0.8773 and a midway profile between both commercial kits with 87.76% sensitivity and 90.48% clinical specificity resulting in a Youden Index of 78.24. This midway trend was further confirmed on 100 double-vaccinated individuals. The BA.2-sVNT performance was similar to that of the Genscript test. Finally, sVNT ability to assess neutralizing antibodies against Omicron BA.5 was validated on a double-vaccinated cohort (n=100) and an Omicron-breakthrough infection cohort (n=91). Correlation analysis revealed a strong association (r=0.8583) between BA.5-sVNT and VNT. Accurate classification was confirmed by receiving operating characteristic (ROC) analysis reporting an area under the curve (AUC) of 0.9543. In conclusion, the sVNT allows for efficient prediction of immune protection against the various VOCs.

NINCO (Out-Of-Distribution detection dataset for ImageNet)

The NINCO (No ImageNet Class Objects) dataset is introduced in the ICML 2023 paper <a href="https://arxiv.org/abs/2306.00826">In or Out? Fixing ImageNet Out-of-Distribution Detection Evaluation</a>. The images in this dataset are free from objects that belong to any of the 1000 classes of ImageNet-1K (ILSVRC2012), which makes NINCO suitable for evaluating out-of-distribution detection on ImageNet-1K . The NINCO main dataset consists of 64 OOD classes with a total of 5879 samples. These OOD classes were selected to have no categorical overlap with any classes of ImageNet-1K. Each sample was inspected individually by the authors to not contain ID objects. Besides NINCO, included are (in the same .tar.gz file) truly OOD versions of 11 popular OOD datasets with in total 2715 OOD samples. Further included are 17 OOD unit-tests, with 400 samples each. Code for loading and evaluating on each of the three datasets is provided at <a href="https://github.com/j-cb/NINCO">https://github.com/j-cb/NINCO</a>.   When using NINCO, please consider citing (besides the bibtex given below) the following data sources that were used to create NINCO: <ul> <li>Hendrycks et al.: ”Scaling out-of-distribution detection for real-world settings”, ICML, 2022.  </li> <li>Bossard et al.: ”Food-101 – mining discriminative components with random forests”, ECCV 2014.  </li> <li>Zhou et al.: ”Places: A 10 million image database for scene recognition”, IEEE PAMI 2017.  </li> <li>Huang et al.: ”Mos: Towards scaling out-of-distribution detection for large semantic space”, CVPR 2021.  </li> <li>Li et al.: ”Caltech 101 (1.0)”, 2022.</li> <li>Ismail et al.: ”MYNursingHome: A fully-labelled image dataset for indoor object classification.”, Data in Brief (V. 32) 2020.</li> <li>The iNaturalist project: https://www.inaturalist.org/</li> </ul> When using NINCO_popular_datasets_subsamples, additionally to the above, please consider citing: <ul> <li>Cimpoi et al.: ”Describing textures in the wild”, CVPR 2014.  </li> <li>Hendrycks et al.: ”Natural adversarial examples”, CVPR 2021.  </li> <li>Wang et al.: ”Vim: Out-of-distribution with virtual-logit matching”, CVPR 2022.  </li> <li>Bendale et al.: ”Towards Open Set Deep Networks”, CVPR 2016.  </li> <li>Vaze et al.: ”Open-set Recognition: a Good Closed-set Classifier is All You Need?”, ICLR 2022.  </li> <li>Wang et al.: ”Partial and Asymmetric Contrastive Learning for Out-of-Distribution Detection in Long-Tailed Recognition.” ICML, 2022.  </li> <li>Galil et al.: “A framework for benchmarking Class-out-of-distribution detection and its application to ImageNet”, ICLR 2023.  </li> </ul> For citing our paper, we would appreciate using the following bibtex entry (this will be updated once the ICML 2023 proceedings are public): @inproceedings{     bitterwolf2023ninco,     title={In or Out? Fixing ImageNet Out-of-Distribution Detection Evaluation},     author={Julian Bitterwolf and Maximilian Mueller and Matthias Hein},     booktitle={ICML},     year={2023},     url={https://proceedings.mlr.press/v202/bitterwolf23a.html} }

Dataset for Paper: Text Line Detection and Recognition of Greek Polytonic Documents

Dataset for Paper: Text Line Detection and Recognition of Greek Polytonic Documents, P. Kaddas, B. Gatos, K. Palaiologos, K. Christopoulou and K. Kritsis, 4th Workshop on Machine Learning (WML), San Jose, California, USA We introduce a new dataset, named GTLD-small dataset, with annotated text line quadrilateral polygons of 1.642 documents, including annotations on 3 datasets (Tobacco-3482, PIOP and ShakeIT dataset) <table> <caption>Overview of the datasets included in this work and the number of images used for training, validation and testing.</caption> <thead> <tr> <th scope="col">Collection</th> <th scope="col">#Total</th> <th scope="col">#train</th> <th scope="col">#val</th> <th scope="col">#test</th> </tr> </thead> <tbody> <tr> <td>PIOP-small</td> <td>950</td> <td>672</td> <td>90</td> <td>188</td> </tr> <tr> <td>ShakeIT-small</td> <td>357</td> <td>264</td> <td>27</td> <td>66</td> </tr> <tr> <td>Tobacco-3482-small</td> <td>335</td> <td>240</td> <td>30</td> <td>65</td> </tr> </tbody> </table>