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.
9,302
datasets available to search
ShareScore release 0.9.0
Dataset results
9,302 results for “DETECT”
Switched capacitor filter transient responses for fault detection
<h2>Dataset Description</h2> <p>This dataset comprises both measured and simulated responses of second-order switched capacitor filters designed for failure detection. The study encompasses two types of filters:</p> <ul> <li> Overdamped</li> <li> Underdamped</li> </ul> <p>The measured data is segregated into two subsets:</p> <h3>Training Data</h3> <p>The training data exclusively contains non-failure responses. It serves as the foundation for model training and validation.</p> <h3>Test Data</h3> <p>The test data includes both failure and non-failure responses. Each response is labeled to indicate whether it is a non-failure or failure response, specifying which capacitor has been modified and its new value.</p> <p>For the simulated data, it is exclusively utilized for testing purposes. For each simulated case (% deviation of the nominal value), a metadata file named "*_metadata.csv" is provided. Each row in this file meticulously outlines the characteristics of the response corresponding to the elements in the accompanying .npy file.</p>
Dataset for the detection and gender classification of crickets Acheta domesticus
<p>The dataset can be used for training AI models for detection and sex classification of Acheta domesticus crickets.<br>The dataset is already divided into training, validation and test dataset. <br>The format is the classical COCO/YOLO format for object detection tasks.<br>The file contain a *.yaml configuration file that allows to define the dataset root directory, the relative paths to training/validation/testing image directories or *.txt files containing image paths, and a dictionary of class names. In this case the class names are:</p> <ul> <li>0: cricket_M (i.e., male crickets);</li> <li>1: cricket_F (i.e., female crickets);</li> </ul> <p>The images are labelled by an expert operator and acquired with the setup described by Giulietti et al. in "Vision Measurement System for Gender-based Counting of Acheta domesticus".<br>Labels for this format should be exported to YOLO format with one *.txt file per image. If there are no objects in an image, no *.txt file is required. The *.txt file should be formatted with one row per object in class x_center y_center width height format. Box coordinates must be in normalized xywh format (from 0 to 1). If your boxes are in pixels, you should divide x_center and width by image width, and y_center and height by image height. Class numbers should be zero-indexed (start with 0).</p>
Balanced Smoking Detection
Open the record for dataset details and reuse information.
Deep transfer learning for detection of breast arterial calcifications on mammograms: a comparative study
<p>Introduction Breast arterial calcifications (BAC) are common incidental findings on routine mammograms, which<br>have been suggested as a sex-specific biomarker of cardiovascular disease (CVD) risk. Previous work showed the efficacy<br>of a pretrained convolutional network (CNN), VCG16, for automatic BAC detection. In this study, we further tested<br>the method by a comparative analysis with other ten CNNs.<br>Material and methods Four-view standard mammography exams from 1,493 women were included in this retrospective<br>study and labeled as BAC or non-BAC by experts. The comparative study was conducted using eleven<br>pretrained convolutional networks (CNNs) with varying depths from five architectures including Xception, VGG,<br>ResNetV2, MobileNet, and DenseNet, fine-tuned for the binary BAC classification task. Performance evaluation<br>involved area under the receiver operating characteristics curve (AUC-ROC) analysis, F1-score (harmonic mean of precision<br>and recall), and generalized gradient-weighted class activation mapping (Grad-CAM++) for visual explanations.<br>Results The dataset exhibited a BAC prevalence of 194/1,493 women (13.0%) and 581/5,972 images (9.7%). Among<br>the retrained models, VGG, MobileNet, and DenseNet demonstrated the most promising results, achieving AUCROCs<br>> 0.70 in both training and independent testing subsets. In terms of testing F1-score, VGG16 ranked first, higher<br>than MobileNet (0.51) and VGG19 (0.46). Qualitative analysis showed that the Grad-CAM++ heatmaps generated<br>by VGG16 consistently outperformed those produced by others, offering a finer-grained and discriminative localization<br>of calcified regions within images.<br>Conclusion Deep transfer learning showed promise in automated BAC detection on mammograms, where relatively<br>shallow networks demonstrated superior performances requiring shorter training times and reduced resources.<br>Relevance statement Deep transfer learning is a promising approach to enhance reporting BAC on mammograms<br>and facilitate developing efficient tools for cardiovascular risk stratification in women, leveraging large-scale mammographic<br>screening programs.<br>Key points<br>• We tested different pretrained convolutional networks (CNNs) for BAC detection on mammograms.<br>• VGG and MobileNet demonstrated promising performances, outperforming their deeper, more complex<br>counterparts.<br>• Visual explanations using Grad-CAM++ highlighted VGG16’s superior performance in localizing BAC.</p>
Flattery Detection from Speech/Text
<p>Features for the dataset presented in "This Paper Had the Smartest Reviewers - Flattery Detection Utilising an Audio-Textual Transformer-Based Approach", accepted at INTERSPEECH 2024. </p> <p>For details, see: </p> <p>Github: <a href="https://github.com/lc0197/flattery_from_speech">https://github.com/lc0197/flattery_from_speech</a></p> <p>ArXiv: TODO</p> <p>The data is available for non-commercial research purposes only. Please note that the raw data is not published due to copyright constraints. </p>
Immunosequencing identifies signatures of anti-tumor T cell responses for early detection of nasopharyngeal carcinoma
<p>We profiled the repertoires of peripheral blood TCRβ chains from 890 individuals including NPC patients, at-risk controls positive for serum EBV VCA-IgA antibody, and seronegative controls through bulk TCRβ sequencing. Library : TCR-seq of genomic DNA extracted from peripheral blood cells.FASTQ files were then processed to extract TCRβ repertoires using MiXCR (v3.0.6)</p>
Single-molecule detection and super-resolution imaging with a portable and adaptable 3D-printed microscopy platform (Brick-MIC)
<p>Supplementary Videos, CAD 3D-printing templates and assembly instructions</p>
Detection and tracking of turbulent structures in the edge of tokamak plasmas using machine learning
<p>This repository contains the data used in the study entitled "Detection and tracking of turbulent structures in the edge of tokamak plasmas: use of an ultra-fast camera and comparison of machine learning and Kalman filter methods". The data was collected using an ultra-fast camera on the COMPASS device.</p> <p><strong><em>This repository includes:</em></strong></p> <p><strong>Fast passive imaging data:</strong> High-frequency captures of turbulent structures present in the edge of tokamak plasmas.<br><strong>Labels :</strong> Annotation of turbulent structures for training and validation of detection and tracking algorithms.<br><strong>Detection and tracking results using YOLO:</strong> Outputs from the YOLO (You Only Look Once) model applied to turbulent data.<br><br><strong><em>Objective</em></strong><br>This dataset is intended to provide a complete and reproducible set for research into the detection and tracking of turbulent structures in tokamak plasmas. It can be used to compare the performance of machine learning methods with traditional techniques, and to encourage further research in this crucial area for plasma physics and nuclear fusion.</p> <p><br>Researchers are encouraged to use these data to:</p> <p>Replicate the results of the original study.<br>Develop and test new techniques for detecting and tracking turbulent structures.Compare the performance of machine learning algorithms with conventional methods.</p> <p><br>References Please cite the repository as follows: S. Chouchene et al., (2024). Detection and tracking of turbulent structures in the edge of tokamak plasmas using machine learning. Zenodo. https://doi.org/10.5281/zenodo.12608068</p>
SH17 Dataset for PPE Detection
<p>We propose <strong>S</strong>afe <strong>H</strong>uman dataset consisting of <strong>17</strong> different objects referred to as <strong>SH17 dataset</strong>. We scrapped images from the <a href="https://pexels.com/" target="_blank" rel="nofollow noopener noreferrer">Pexels website</a>, which offers <a href="https://www.pexels.com/license/" target="_blank" rel="nofollow noopener noreferrer">clear usage rights</a> for all its images, showcasing a range of human activities across diverse industrial operations.</p> <p>To extract relevant images, we used multiple queries such as manufacturing worker, industrial worker, human worker, labor, etc. The tags associated with Pexels images proved reasonably accurate. After removing duplicate samples, we obtained a dataset of 8,099 images. The dataset exhibits significant diversity, representing manufacturing environments globally, thus minimizing potential regional or racial biases. Samples of the dataset are shown below.</p> <h3>Key features</h3> <ul> <li>Collected from diverse industrial environments globally</li> <li>High quality images (max resolution 8192x5462, min 1920x1002)</li> <li>Average of 9.38 instances per image</li> <li>Includes small objects like ears and earmuffs (39,764 annotations < 1% image area, 59,025 annotations < 5% area)</li> </ul> <h2>Classes</h2> <ol> <li>Person</li> <li>Head</li> <li>Face</li> <li>Glasses</li> <li>Face-mask-medical</li> <li>Face-guard</li> <li>Ear</li> <li>Earmuffs</li> <li>Hands</li> <li>Gloves</li> <li>Foot</li> <li>Shoes</li> <li>Safety-vest</li> <li>Tools</li> <li>Helmet</li> <li>Medical-suit</li> <li>Safety-suit</li> </ol> <p>The data consists of three folders,</p> <ul> <li><strong>images</strong> contains all images</li> <li><strong>labels</strong> contains labels in YOLO format for all images</li> <li><strong>voc_labels</strong> contains labels in VOC format for all images</li> <li><strong>train_files.txt</strong> contains list of all images we used for training</li> <li><strong>val_files.txt</strong> contains list of all images we used for validation</li> </ul> <h2>Disclaimer and Responsible Use:</h2> <p>This dataset, scrapped through the Pexels website, is intended for educational, research, and analysis purposes only. You may be able to use the data for training of the Machine learning models only. Users are urged to use this data responsibly, ethically, and within the bounds of legal stipulations.</p> <h2>Users should adhere to <a href="https://www.pexels.com/license/" target="_blank" rel="nofollow noopener noreferrer">Copyright Notice of Pexels </a>when utilizing this dataset.</h2> <p>Legal Simplicity: All photos and videos on Pexels can be downloaded and used for free.</p> <h4>Allowed 👌</h4> <ul> <li>All photos and videos on Pexels are free to use.</li> <li>Attribution is not required. Giving credit to the photographer or Pexels is not necessary but always appreciated.</li> <li>You can modify the photos and videos from Pexels. Be creative and edit them as you like.</li> </ul> <h4>Not allowed 👎</h4> <ul> <li>Identifiable people may not appear in a bad light or in a way that is offensive.</li> <li>Don't sell unaltered copies of a photo or video, e.g. as a poster, print or on a physical product without modifying it first.</li> <li>Don't imply endorsement of your product by people or brands on the imagery.</li> <li>Don't redistribute or sell the photos and videos on other stock photo or wallpaper platforms.</li> <li>Don't use the photos or videos as part of your trade-mark, design-mark, trade-name, business name or service mark.</li> </ul> <h4>No Warranty Disclaimer:</h4> <p>The dataset is provided "as is," without warranty, and the creator disclaims any legal liability for its use by others.</p> <h4>Ethical Use:</h4> <p>Users are encouraged to consider the ethical implications of their analyses and the potential impact on broader community.</p> <h4>GitHub Page:</h4> <p><a href="https://github.com/ahmadmughees/SH17dataset" target="_blank" rel="nofollow noopener noreferrer">https://github.com/ahmadmughees/SH17dataset</a></p>
Stimuli for "Human Detection of Political Speech Deepfakes across Transcripts, Audio, and Video"
<p>This dataset contains all stimuli used in "Human Detection of Political Speech Deepfakes across Transcripts, Audio, and Video" by Matthew Groh, Aruna Sankaranarayanan, Nikhil Singh, Dong Young Kim, Andrew Lippman, and Rosalind Picard.</p> <p>The videos are contained in the Materials folder and the Data Folder provides two .csv files that map the video filenames with their metadata. </p> <p>For video sources, see Table 19 in the arXiv version of the paper: https://arxiv.org/abs/2202.12883</p>
Malicious Large Language Models Detection using Metadata Information
<p># Introduction<br># This is the replication package for the paper "Malicious Large Language Models Detection using Metadata Information".</p> <p><br>## Task Definition<br>Given the information of LLM, the task is to identify whether it is a malicious LLM that may attack software systems. We treat the task as binary classification (0/1), where 1 stands for malicious LLMs and 0 for malicious-free LLMs.</p> <p><br>## The static of dataset<br>### Data Format<br>Before preprocessing dataset, each line in the uncompressed file represents multiple metadata of one large language model (LLM). One row is illustrated below.<br> - **idx:** the index of example<br> - **repo_id:** the id of LLM (e.g., microsoft/codebert-base)<br> - **tags:** the tags of LLM<br> - **pipeline_tag:** the pipeline_tag of LLM<br> - **downloads:** the number of downloads<br> - **created_time:** the created time of LLM<br> - **modelCard:** the text content of LLM<br> - **num_discussion:** the number of discussions<br> - **discussion:** the discussions of LLM<br> - **para_size:** the size of LLM<br> - **tensor_type:** the type of LLM<br> - **num_commit:** the number of commits<br> - **commit:** the commit of LLM</p> <p>After preprocessing dataset, you can obtain three .csv files, i.e. train.csv, valid.csv, test.csv<br> - **idx:** the index of example.<br> - **repo_id:** the id of LLM (e.g., microsoft/codebert-base).<br> - **tags:** the tags of LLM.<br> - **pipeline_tags:** the pipeline_tag of LLM.<br> - **created_time:** the created time of LLM.<br> - **model_size:** the size of LLM.<br> - **Tensor_type:** the type of LLM.<br> - **is_model_card:** Whether to include model card. If the model has model card, the value is 1, otherwise, the value is 0.<br> - **malicious_model_card:** Whether to include keywords describing the malicious model in model card. If the model card has malicious keywords, the value is 1, otherwise, the value is 0<br> - **repository_link:** Whether to include repository link: GitHub link, Arxiv link, homepage link, bugs link and issues link in model card. If the model card has link, the value is 1, otherwise, the value is 0.<br> - **dataset_info:** Whether to include the adopted dataset information in model card. If the model card has dataset information, the value is 1, otherwise, the value is 0.<br> - **metrics_info:** Whether to include the evaluated metrics information in model card. If the model card has evaluation metrics information, the value is 1, otherwise, the value is 0.<br> ?- **script_info:** Whether to include script information. If the model has script information, the value is 1, otherwise, the value is 0.<br> - **config_content:** The content of the configuration script file. This value is string type.<br> - **stakeholder_name:** The name of authors, contributors, and maintainers. This value is string type.<br> - **number_discussion:** The number of discussion. <br> - **num_pr:** The number of pull request.<br> - **malicious_discussion:** Whether the discussion contains malicious behavior keywords. If the discussion has malicious behavior keywords, the value is 1, otherwise, the value is 0.<br> - **number_commit:** The number of commit.<br> - **malicious_commit:** Whether the title and message of commits contain malicious behavior keywords. If the commit has malicious behavior keywords, the value is 1, otherwise, the value is 0.<br> - **z_download:** The z-score of number of download.<br> - **z_like:** The z-score of number of likes.</p> <p><br>### Data Statistics<br>Data statistics of the dataset are shown in the below table:</p> <p>| #File Names | #Examples |<br>| ------------------- | :------------------------: |<br>| dataset_feature.csv | 578,502 (560,257/18,245) | <br>| train_imbalance.csv | 462,801 (448205/14596) |<br>| valid_imbalance.csv | 57,849 (56025/1824) |<br>| test_imbalance.csv | 57,852 (56027/1825) |<br>| train_balance.csv | 29192 (14596/14596) |<br>| valid_balance.csv | 3648 (1824/1824) |<br>| test_balance.csv | 3650 (1825/1825) |<br>| train_imbalance_50.csv | 289252 (280129/9123) |<br>| train_imbalance_60.csv | 347101 (336154/10947) |<br>| train_imbalance_70.csv | 404952 (392180/12772) |</p> <p>29646 models contain github link<br>all_dataset.csv 596383<br>all_dataset_information.csv 589140 (safe_dataset.csv(570549), unsafe_dataset.csv(18591))<br>safe_dataset_information.csv (559582), unsafe_dataset_information.csv (18212)</p> <p>Description Feature: 'malicious_model_card', 'repository_link', 'dataset_info', 'metrics_info', 'config_conteng'<br>Stakeholder Feature: 'stakeholder_name'<br>Event Feature: 'num_pr', 'number_commit', 'malicious_commit',<br>Context Feature: 'z_download', 'z_like'</p> <p><br>## Pipeline-MPTMHunter<br>We also provide a pipeline that fine-tunes [MPTMHunter](https://doi.org/10.5281/zenodo.12578531) on this task.</p> <p>### Experimental environment configuration<br>```bash<br>huggingface_hub 0.23.1<br>libxgboost 2.0.3<br>lightgbm 4.3.0<br>networkx 3.2.1<br>nltk 3.8.1<br>numpy 1.26.3<br>openssl 3.0.13<br>pandas 2.2.1<br>pillow 10.2.0<br>scikit-learn 1.4.2<br>scipy 1.13.0<br>torch 2.3.0+cu118<br>torchaudio 2.3.0+cu118<br>torchvision 0.18.0+cu118<br>tqdm 4.66.2<br>transformers 4.37.2<br>xgboost 2.0.3<br>```</p> <p>### Dataset Collection Script<br>```bash<br>python ./script/DataExtraction.ipynb<br>python ./script/dataset_spider.py<br>python ./script/config_crawl.py<br>```</p> <p>### Dataset Preprocess Script<br>```bash<br>python feature_generation.py --input_file='../dataset/dataset_information.csv' --output_file='../dataset/dataset_feature.csv'</p> <p>python feature_generation.py --input_file='../dataset/real_world_dataset_information_0701.csv' --output_file='../dataset/real_world_dataset_feature_0701.csv'<br>```</p> <p>### Model Training Script<br>```bash<br>python run_codet5_lstm.py --output_dir='../saved_models/codet5_lstm_imbalance_final' --model_type=codet5 --tokenizer_name='../models/codet5' --model_name_or_path='../models/codet5' --do_train --train_data_file='../dataset/train_imbalance_70.csv' --eval_data_file='../dataset/valid_imbalance_70.csv' --test_data_file='../dataset/test_imbalance.csv' --epoch=3 --block_size=510 --train_batch_size=64 --eval_batch_size=64 --learning_rate=2e-5 --max_grad_norm=1.0 --evaluate_during_training --seed=123456<br>```</p> <p>### Model Inference Script<br>```bash<br>python run_codet5_lstm.py --output_dir='../saved_models/codet5_lstm_imbalance' --model_type=codet5 --tokenizer_name='../models/codet5' --model_name_or_path='../models/codet5' --do_eval --do_test --train_data_file='../dataset/train_imbalance_70.csv' --eval_data_file='../dataset/valid_imbalance_70.csv' --test_data_file='../dataset/test_imbalance.csv' --epoch=3 --block_size=510 --train_batch_size=64 --eval_batch_size=64 --learning_rate=2e-5 --max_grad_norm=1.0 --evaluate_during_training --seed=123456<br>```</p> <p>### Evaluation Script<br>```bash<br>python ../evaluation/evaluation.py -a ../dataset/test_balance.csv -p ../saved_models/codebert_imbalance_all/predictions.txt<br>python ../evaluation/evaluation.py -a ../dataset/test_imbalance.csv -p ../dataset/predictions.txt<br>```</p> <p>## Result<br>The results on the test set are shown as below (We use the OpenTextClassification as the baseline):</p> <p>| Methods | ACC | Precision | Recall | F1-Score |<br>| Random Forest | 97.18% | 83.80% | 13.04% | 22.57% |<br>| LR | 96.80% | 43.46% | 4.55% | 8.23% |<br>| LightGBM | 97.62% | 95.18% | 25.97% | 40.81% |<br>| TextRNN | 96.86% | 73.33% | 0.60% | 1.20% |<br>| TextCNN | 98.89% | 95.61% | 68.00% | 79.47% |<br>| TextRCNN | 98.87% | 95.36% | 67.62% | 79.13% |<br>| TextRNN_Att | 98.94% | 95.29% | 69.86% | 80.62% |<br>| MPTMHunter | ** 99.99% ** | ** 99.95% ** | ** 99.78% ** | ** 99.86% ** |</p>
Evaluation of minimum water-filled pore space (WFPS) to ensure detectable N2 fluxes for Central Experiments in DASIM
<p><strong>Methods</strong></p><p>Incubation of DASIM soils in Thünen microcosm system under the following conditions:</p><ul><li>N2-depleted atmosphere(2% N2 in pure He or in 20%O2/78%He) (Lewicka-Szczebak et al., 2017)</li><li>Target flow rate: 20 mL/min</li><li>Core diameter: 14,4 cm</li><li>Core heights:10 cm</li><li>Bulk density: according to DASIM protocol: Fuhrberg, arable (sand) = 1.5; Rottalmünster, arable (silt loam) = 1.3; Gießen, grassland = clay loam (1.2)</li><li>15N-labeling: addition of K15NO3 (60 at% 15N), target content of 50 mg N/kg</li><li>Litter: all treatments <strong>without</strong> litter</li><li>WFPS levels: 60 /65/70/75 %</li><li>Incubation sequence: 1 week aerobic (20%O2), one week anaerobic</li><li>Temperature: 20°C</li><li>Gas sampling for IRMS: twice a week</li><li>GC analysis online about 3 times a day</li></ul><p>IRMS analysis: see Lewicka-Szczebak et al. 2013)</p><p>Calculation of ap and fs: see Spott et al. (2006) and Buchen et al. (2016)</p><p>Determination of limit of detection (LOD): Monte-Carlo modelling using SD of 29R and 30R of IRMS analysis</p><p>soils were sieved to 10 mm, preincubates at 50% WHCmax and 20°C for two weeks, fertilized, packed to target height and BD and wetted to target WFPS. Final WFPS and NO3--N was not measured.</p><p> </p>
Dataset of the paper "Minimal detectable change of gait and balance measures in older neurological patients: estimating the standard error of the measurement from before-after rehabilitation data thanks to the linear mixed-effects models"
<p>The Excel file contains the dataset whose analysis has been presented in the manuscript entitled <em>"Minimal detectable change of gait and balance measures in older neurological patients: estimating the standard error of the measurement from before-after rehabilitation data thanks to the linear mixed-effects models" </em>and published in J Neuroeng Rehabil (doi: 10.1186/s12984-024-01339-4; PMID: 38566189).</p> <p>This data set cannot be publicly available because of sensitive information. Please send your request to: a.caronni@auxologico.it.</p>
Dataset for Leakage Detection Index method
<p>Leakage detection index (LDI) technique is used to determine the presence of leakage through a temporal comparison of the cross-spectral density (CSD) of pipeline acceleration response measured at multiple locations along the pipeline length. </p>
singing voice detection dataset ALL-Pub-SVD-In-One
<p>singing voice detection datasets:</p> <p>it contains the public dataset of Jamendo Dataset, and the RWC dataset , and the iKala dataset, and the MIR-1K dataset , and teh MedlyDB dataset.</p> <p>Jamendo are same with the raw.</p> <p>RWC use the simply mode same with the lable in Jamendo just sing or nosing.</p> <p>iKala and MIR-1K are full wavfiles with the generated lables from the PitchLables. 0 for nosing and other for sing.</p> <p>MedlyDB use part wavfiles from the RAW dataset and also generated lables from the MELODY_2 lables with 0 for nosing and other for nosing.</p> <p> </p>
Deep Learning for Real Time 3D Multi-Object Detection, Localization, and Tracking : Application to Smart Mobility
<p>In core computer vision tasks, we have witnessed significant advances in object detection, localisation and tracking. However, there are currently no methods to detect, localize and track objects in road environments, and taking into account real-time constraints. In this paper, our objective is to develop a deep learning multi object detection and tracking technique applied to road smart mobility. Firstly, we propose an effective detector-based on YOLOV3 [1] which we adapt to our context. Subsequently, to localize successfully the detected objects, we put forward an adaptive method aiming to extract 3D information, i.e., depth maps. To do so, a comparative study is carried out taking into account two approaches: Monodepth2 [2,3] for monocular vision and MADNEt [4] for stereoscopic vision. These approaches are then evaluated over datasets containing depth information in order to discern the best solution that performs better in real-time condition. Object tracking is necessary in order to mitigate the risks of collisions. Unlike, traditional tracking approaches which requires target initialization beforehand, our approach consists of using information from object detection and distance estimation to initialize targets and to track them later. Expressly, we propose here to improve SORT [5] approach for 3D object tracking. We introduce an extended Kalman filter [6] to better estimate the position of objects. Extensive experiments carried out on KITTI dataset [7] prove that our proposal outperforms state-of-the-art approches. </p>
VoiceWukong: Benchmarking Deepfake Voice Detection (part_ab)
<h1>VoiceWukong</h1> <p><em><strong>VoiceWukong </strong>is a comprehensive benchmark for deepfake voice detection, designed to evaluate the performance of various detectors in real-world application scenarios.</em></p> <p><strong>Dataset Features</strong></p> <ul> <li>Large Scale: Contains 265,200 English and 148,200 Chinese deepfake voice samples</li> <li>Diverse Sources: Covers voice samples generated by 19 commercial tools and 15 open-source tools</li> <li>Real-world Scenarios: Constructed 38 data variants covering 6 types of audio manipulations common in practical applications</li> <li>Bilingual Support: Supports evaluation in both Chinese and English languages</li> </ul> <p><strong>Evaluation Results</strong></p> <ul> <li>Conducted comprehensive evaluations on 12 state-of-the-art deepfake voice detectors</li> <li><a href="https://github.com/TakHemlata/SSL_Anti-spoofing">AASIST2</a> achieved the best performance with an Equal Error Rate (EER) of 13.50%</li> <li>Other detectors showed EERs exceeding 20%</li> <li>Results indicate significant challenges for current detectors in practical applications</li> </ul> <p><strong>Human-Machine Comparison Study</strong></p> <ul> <li>Conducted user studies with over 300 participants</li> <li>Comparative analysis of detection capabilities among humans, detectors, and multimodal large language models (<a href="https://github.com/QwenLM/Qwen2-Audio">Qwen2-Audio</a>)</li> <li>Different detectors and humans showed varying identification capabilities for deepfake voices at different deception levels</li> <li>Multimodal large language models demonstrated no effective detection ability</li> </ul> <h2>Dataset</h2> <p>This is the second part of the dataset, and it requires the complete download of both <a href="https://zenodo.org/records/13731918"><strong><em>part_aa</em></strong></a> and <a href="https://zenodo.org/records/13732412"><strong><em>part_ab</em></strong></a> for proper extraction and use. Please ensure that both files are in the same folder. For a detailed introduction to the data, please refer to our paper (to be made available).</p> <p>The first part (part_aa) is at <a href="../records/13731918">part_aa</a><br>extract command : <code>cat VoiceWukong.part_* | tar -xz</code></p> <h2>Leaderboard</h2> <p>Our leaderboard presents comprehensive evaluation results in three main sections:</p> <ol> <li><strong>Overall Performance</strong> - General evaluation metrics for each detector across the entire dataset, providing a broad view of detection capabilities.</li> <li><strong>Manipulation-specific Performance</strong> - Detailed results showing how each detector performs under different types of audio manipulations, offering insights into specific strengths and weaknesses.</li> <li><strong>User Study-based Evaluation</strong> - Performance analysis of detectors on deepfake voices categorized by difficulty levels based on our user study results, demonstrating detector effectiveness across varying deception capabilities.</li> </ol> <p>Visit our <a href="https://voicewukong.github.io/">leaderboard(github.io)</a> for detailed performance metrics and rankings. Additionally, we provide a copy of the leaderboard code <a href="https://zenodo.org/uploads/14650793">here</a> for permanent storage.</p> <div> <h2>Evaluated Detectors' Weighted Models</h2> <ul> <li>All evaluated detectors’ weighted models can be obtained from <a title="https://huggingface.co/VoiceWukong/VoiceWukong/" href="https://huggingface.co/VoiceWukong/VoiceWukong/">huggingface.co</a>. Additionally, we provide a copy of the weights files <a href="https://zenodo.org/uploads/14650793">here</a> for premanent storage.</li> </ul> <h2>User Study Results & Original Outputs</h2> <ul> <li>This <a href="https://github.com/VoiceWukong/VoiceWukong">code repository(github)</a> stores our <a href="https://github.com/VoiceWukong/VoiceWukong/tree/main/Userstudy/result">user study results</a> and the <a href="https://github.com/VoiceWukong/VoiceWukong/tree/main/OutputScore">original outputs</a> of the evaluation detectors. Additionally, we provide a copy of the code repository <a href="https://zenodo.org/uploads/14650793">here</a> for permanent storage.</li> </ul> </div> <p>Note: <em><strong>VoiceWukong</strong></em> <strong>prohibits use for <em>commercial purposes.</em></strong></p>
Fundus dataset for detection of Diabetic Retinopathy
<p><span>Fundus images with dimensions 3680x3288 collected at Natasha Eye Care and Research Institute in Pune, Maharashtra, India. The Symbiosis Institutional Ethics Committee (BHR) SIU granted authorization for the project under approval code SIU/IEC/583. The images were captured using a nonmydriatic as per the protocol, and Fundus images using an Eidon machine. A Fundus scan requires 1-2 minutes if the patient is cooperative. Two ophthalmologists then annotated the images. ETDRS scale is used for the classification of images. A signed consent form was given by every patient. This dataset can be used by researchers and doctors to develop automated diagnostic tools for early detection of diabetic retinopathy (DR). </span></p> <p> </p>
XDF: A Large-Scale Dataset for Evaluating Video Deepfake Detection Across Multiple Manipulation Techniques
Open the record for dataset details and reuse information.
PAN21 Authorship Analysis: Style Change Detection
<p>This is the dataset for the <a href="https://pan.webis.de/clef21/pan21-web/style-change-detection.html">Style Change Detection</a> task of PAN 2021.</p> <p>The goal of the style change detection task is to identify text positions within a given multi-author document at which the author switches. </p> <p><strong>Tasks</strong></p> <p>Given a document, we ask participants to answer the following three questions:</p> <ul> <li><em>Single vs. Multiple.</em> Given a text, find out whether the text is written by a single author or by multiple authors (task 1).</li> <li><em>Style Change Basic.</em> Given a text written by two or more authors and that contains a number of style changes, find the position of the changes (task 2).</li> <li><em>Style Change Real-World.</em> Given a text written by two or more authors, find all positions of writing style change, i.e., assign all paragraphs of the text uniquely to some author out of the number of authors you assume for the multi-author document (task 3).</li> </ul> <p>All documents are provided in English and may contain an arbitrary number of style changes, resulting from at most five different authors. However, style changes may only occur between paragraphs (i.e., a single paragraph is always authored by a single author and does not contain any style changes).</p> <p><strong>Data</strong></p> <p>The dataset is split into three parts:</p> <ol> <li><em>training set:</em> Contains 70% of the whole data set and includes ground truth data. Use this set to develop and train your models.</li> <li><em>validation set:</em> Contains 15% of the whole data set and includes ground truth data. Use this set to evaluate and optimize your models.</li> <li><em>test set:</em> Contains 15% of the whole data set. For the documents on the test set, you are not given ground truth data. This set is used for evaluation.</li> </ol> <p>The dataset is based on user posts from various sites of the StackExchange network, covering different topics. We refer to each input problem (i.e., the document for which to detect style changes) by an ID, which is subsequently also used to identify the submitted solution to this input problem. We provide one folder for train, validation, and test data.</p> <p>For each problem instance <code>X</code> (i.e., each input document), two files are provided:</p> <ol> <li><code>problem-X.txt</code> contains the actual text, where paragraphs are denoted by <code>\n\n</code>.</li> <li><code>truth-problem-X.json</code> contains the ground truth, i.e., the correct solution in JSON format: <pre><code class="language-json">{ "authors": NUMBER_OF_AUTHORS, "site": SOURCE_SITE, "multi-author": RESULT_TASK1, "changes": RESULT_ARRAY_TASK2, "paragraph-authors": RESULT_ARRAY_TASK3 }</code></pre> The result for task 1 (key "multi-author") is a binary value (1 if the document is multi-authored, 0 if the document is single-authored). The result for task 2 (key "changes") is represented as an array, holding a binary for each pair of consecutive paragraphs within the document (0 if there was no style change, 1 if there was a style change). If the document is single-authored, the solution to task 2 is an array filled with 0s. For task 3 (key "paragraph-authors"), the result is the order of authors contained in the document (e.g., <code>[1, 2, 1]</code> for a two-author document), where the first author is "1", the second author appearing in the document is referred to as "2", etc. Furthermore, we provide the total number of authors and the Stackoverflow site the texts were extracted from (i.e., topic).<br> <br> An example of a multi-author document, where there was a style change between the third and fourth paragraph could look as follows (we only list the relevant key/value pairs here): <pre><code class="language-json">{ "multi-author": 1, "changes": [0,0,1,...], "paragraph-authors": [1,1,1,2,...] }</code></pre> A single-author document would have the following form (again, only listing the relevant key/value pairs): <pre><code class="language-json">{ "multi-author": 0, "changes": [0,0,0,...], "paragraph-authors": [1,1,1,...] }</code></pre> <p> </p> </li> </ol>
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.