Find research datasets worth reusing

opencc-by-4.0Dec 2021View details →

Hand Washing Video Dataset Annotated According to the World Health Organization's Handwashing Guidelines - METC Subset

Overview: This is a lab-based dataset with videos recording volunteers (medical students) washing their hands as part of a hand-washing monitoring and feedback experiment. The dataset is collected in the Medical Education Technology Center (METC) of Riga Stradins University, Riga, Latvia. In total, 72 participants took part in the experiments, each washing their hands three times, in a randomized order, going through three different hand-washing feedback approaches (user interfaces of a mobile app). The data was annotated in real time by a human operator, in order to give the experiment participants real-time feedback on their performance. There are 212 hand washing episodes in total, each of which is annotated by a single person. The annotations classify the washing movements according to the World Health Organization's (WHO) guidelines by marking each frame in each video with a certain movement code. This dataset is part on three dataset series all following the same format: <ul> <li><a href="https://zenodo.org/record/4537209">https://zenodo.org/record/4537209 </a>- data collected in Pauls Stradins Clinical University Hospital</li> <li><a href="https://zenodo.org/record/5808764">https://zenodo.org/record/5808764</a> - data collected in Jurmala Hospital</li> <li><a href="https://zenodo.org/record/5808789">https://zenodo.org/record/5808789</a> - data collected in the Medical Education Technology Center (METC) of Riga Stradins University</li> </ul> Note #1: we recommend that when using this dataset for machine learning, allowances are made for the reaction speed of the human operator labeling the data. For example, the annotations can be expected to be incorrect a short while after the person in the video switches their washing movements. Application: The intention of this dataset is to serve as a basis for training machine learning classifiers for automated hand washing movement recognition and quality control. Statistics: <ul> <li>Frame rate: ~16 FPS (slightly variable, as the video are reconstructed from a sequence of jpg images taken with max framerate supported by the capturing devices).</li> <li>Resolution: 640x480</li> <li>Number of videos: 212</li> <li>Number of annotation files: 212</li> </ul> Movement codes (in JSON files): <ul> <li>1: Hand washing movement — Palm to palm</li> <li>2: Hand washing movement — Palm over dorsum, fingers interlaced</li> <li>3: Hand washing movement — Palm to palm, fingers interlaced</li> <li>4: Hand washing movement — Backs of fingers to opposing palm, fingers interlocked</li> <li>5: Hand washing movement — Rotational rubbing of the thumb</li> <li>6: Hand washing movement — Fingertips to palm</li> <li>0: Other hand washing movement</li> </ul> Note #2: The original dataset of JPG images is available upon request. There are 13 annotation classes in the original dataset: for each of the six washing movements defined by the WHO, "correct" and "incorrect" execution is market with two different labels. In this published dataset, all incorrect executions are marked with code 0, as "other" washing movement. Acknowledgments: The dataset collection was funded by the Latvian Council of Science project: "Automated hand washing quality control and quality evaluation system with real-time feedback", No: lzp - Nr. 2020/2-0309. References: For more detailed information, see this article, describing a similar dataset collected in a different project: <ul> <li> M. Lulla, A. Rutkovskis, A. Slavinska, A. Vilde, A. Gromova, M. Ivanovs, A. Skadins, R. Kadikis, A. Elsts. Hand-Washing Video Dataset Annotated According to the World Health Organization’s Hand-Washing Guidelines. Data. 2021; 6(4):38. <a href="https://doi.org/10.3390/data6040038">https://doi.org/10.3390/data6040038</a> </li> </ul> Contact information: atis.elsts@edi.lv

opencc-by-4.0Dec 2021View details →

Hand Washing Video Dataset Annotated According to the World Health Organization's Handwashing Guidelines - Jurmala Hospital Subset

Overview: This is a large-scale real-world dataset with videos recording medical staff washing their hands as part of their normal job duties in the Jurmala Hospital located in Jurmala, Latvia. There are 2427 hand washing episodes in total, almost all of which are annotated by two persons. The annotations classify the washing movements according to the World Health Organization's (WHO) guidelines by marking each frame in each video with a certain movement code. This dataset is part on three dataset series all following the same format: <ul> <li><a href="https://zenodo.org/record/4537209">https://zenodo.org/record/4537209</a> - data collected in Pauls Stradins Clinical University Hospital</li> <li><a href="https://zenodo.org/record/5808764">https://zenodo.org/record/5808764</a> - data collected in Jurmala Hospital</li> <li><a href="https://zenodo.org/record/5808789">https://zenodo.org/record/5808789</a> - data collected in the Medical Education Technology Center (METC) of Riga Stradins University</li> </ul> Applications: The intention of this dataset is twofold: to serve as a basis for training machine learning classifiers for automated hand washing movement recognition and quality control, and to allow to investigate the real-world quality of washing performed by working medical staff. Statistics: <ul> <li>Frame rate: 30 FPS</li> <li>Resolution: 320x240 and 640x480</li> <li>Number of videos: 2427</li> <li>Number of annotation files: 4818</li> </ul> Movement codes (both in CSV and JSON files): <ul> <li>1: Hand washing movement — Palm to palm</li> <li>2: Hand washing movement — Palm over dorsum, fingers interlaced</li> <li>3: Hand washing movement — Palm to palm, fingers interlaced</li> <li>4: Hand washing movement — Backs of fingers to opposing palm, fingers interlocked</li> <li>5: Hand washing movement — Rotational rubbing of the thumb</li> <li>6: Hand washing movement — Fingertips to palm</li> <li>7: Turning off the faucet with a paper towel</li> <li>0: Other hand washing movement</li> </ul> Acknowledgments: The dataset collection was funded by the Latvian Council of Science project: "Automated hand washing quality control and quality evaluation system with real-time feedback", No: lzp - Nr. 2020/2-0309. References: For more detailed information, see this article, describing a similar dataset collected in a different project: <ul> <li> M. Lulla, A. Rutkovskis, A. Slavinska, A. Vilde, A. Gromova, M. Ivanovs, A. Skadins, R. Kadikis, A. Elsts. Hand-Washing Video Dataset Annotated According to the World Health Organization’s Hand-Washing Guidelines. Data. 2021; 6(4):38. <a href="https://doi.org/10.3390/data6040038">https://doi.org/10.3390/data6040038</a> </li> </ul> Contact information: atis.elsts@edi.lv

opencc-by-4.0Feb 2022View details →

Wikipedia video games similarity dataset with expert annotations

A video games NLP dataset extracted from Wikipedia. For all articles, the figures and tables have been filtered out, as well as the categories and "see also" sections. The article structure, and particularly the sub-titles and paragraphs are kept in these picese. Provided as well are 90 seeds with recommended articles, annotated by human experts.

opencc-by-4.0Dec 2021View details →

Dataset of UAV thermal video sequences with annotations for MOTS benchmarking

Instance segmentation dataset created for the research 'Monitoring Mammalian Herbivores via Convolutional Neural Networks implemented on Thermal UAV imagery'. It comprises 959 frames, 20.647 masks, and 239 tracks, and consists of 7 video sequences depicting aerial thermal imagery of cattle collected with a UAV (Parrot ANAFI Thermal) in two outdoor farms in the Netherlands. Data were acquired at three temperatures (10ºC, 19ºC, and 26.5ºC), under sunny and overcast weather conditions, at various angles of inclination (including nadir), and at heights ranging between 8-28 meters. Ground truth was labeled manually with the Computer Vision Annotation Tool CVAT.

dryad40/100

Common guillemots in the Baltic Sea studied with video surveillance and object detection: raw data, annotations, model, and model outputs

The data comes from common guillemots studied at Stora Karlsö, Sweden between 2019 and 2021. The common guillemots breed at an artificial cliff, and has been filmed continusly from above over three breeding seasons. Using the video material, a YOLOv5 model has been trained to detect adult birds, chicks and eggs. The dataset contains annotations (bounding boxes) used for training the model, the model itself, and outputs from the model (object detections). The data can be used and shared freely.

opencc-zeroSep 2022View details →

opencc-by-4.0Oct 2018View details →

Arrangement of J.S. Bach's Invention No. 13 in A minor (BWV784) for the DIMI-A synthesizer by Erkki Kurenniemi (1970): An annotated video of the master tape

The video presents an annotated analysis of the master tape of the arrangement of J.S.Bach's invention in a minor (BWV784) for DIMI-A synthesizer (1970) by Erkki Kurenniemi. Upload also includes picture and excel files of analysis material. Analysis is partially published in Lassfolk, K., Suominen, J., & Ojanen, M. (2015). Interaction of Music and Technology: The Music and Musical Instruments of Erkki Kurenniemi. In J. Krysa, & J. Parikka (Eds.), Writing and Unwriting (Media) Art History : Erkki Kurenniemi in 2048 (pp. 261-277). [19] (Leonardo Book Series). Cambridge, Mass.: MIT Press. The video is also available for streaming/embedding via https://vimeo.com/278832133

opencc-by-4.0Apr 2023View details →

Annotated Video Dataset of Fencing Movements and Corresponding Error Patterns

Video dataset containing 1289 videos and their augmentation of four fencing movements, performed by a variety of fencers. The main actions included are counterattack, lunge, flèche and parry. For each movement samples with typical error patterns are included and annotated. Additionally labels for mulit-labelling are documented. The corresponding paper "Mastering Fencing Techniques with Machine Learning: A Video-Based Classification and Correction System" is published at the 10th IEEE Swiss Conference on Data Science (SDS 2023)

dryad40/100

Common guillemots in the Baltic Sea studied with video surveillance and object detection: raw data, annotations, model, and model outputs

Open the record for dataset details and reuse information.

publicDec 2022View details →

opencc-by-4.0Dec 2019View details →

Video footage of vessel annotation framework

This video illustrates our vessel annotation framework. This is part of a paper submission. Once the paper is accepted, this information will be updated.

opencc-by-4.0Nov 2021View details →

CEAP-360VR: A Continuous Physiological and Behavioral Emotion Annotation Dataset for 360 VR Videos

<a href="http://ieeexplore.ieee.org/document/9599346">CEAP-360VR: A Continuous Physiological and Behavioral Emotion Annotation Dataset for 360° Videos</a> ## General Information We develop the <a href="http://www.dis.cwi.nl/ceap-360vr-dataset/">CEAP-360VR</a> dataset to address the lack of continuously annotated behavioral and physiological datasets for 360 video VR affective computing. Accordingly, this dataset contains a) questionnaires (SSQ, IPQ, NASA-TLX); b) continuous valence-arousal annotations; c) head and eye movements as well as left and right eye pupil diameters while watching videos; d) peripheral physiological responses (ACC, EDA, SKT, BVP, HR, IBI). Our dataset also concludes the data pre-processing, data validating scripts, along with dataset description and key steps in the stage of data acquisition and pre-processing. ## Dataset Structure The  CEAP-360VR folder contains the following six subfolders 1_Stimuli 2_QuestionnaireData 3_AnnotationData 4_BehaviorData 5_PhysioData 6_Scripts The following is a detailed description of each sub-file: 1_Stimuli     - VideoThumbNails         contains the eight thumbNails for each video (.jpg)     - VideoInfo.json         contains the detailed information for eight videos 2_QuestionnaireData     - PXX_Questionnaire_Data.json (X = 1, 2, ..., 32)         contains questionnaire data for each participant 3_AnnotationData     - Raw         contains the raw annotation data captured from the Joy-Con joystick for each participant     - Transformed         contains the transformed valence-arousal data generated from the raw data for each participant     - Frame         contains the re-sampled annotation data from the transformed data for each participant 4_BehaviorData     - Raw         contains the raw behavior data captured from the HTC VIVE Pro Eye Tobii Device for each participant     - Transformed         contains the transformed heam/eye movement data (pitch/yaw) generated from the raw data, as well as pupil diameter data for each participant     - Frame         contains the re-sampled behavior data generated from the transformed data for each participant     - HM_ScanPath         contains the head scanpath data generated from the transformed data for each participant     - EM_Fixation         contains the eye gaze fixation data generated from the transformed data for each participant 5_PhysioData     - Raw         contains the raw physiological data captured from the Empatica E4 wristband for each participant     - Transformed         contains the transformed physiological data generated from the raw data for each participant     - Frame         contains the re-sampled physiological data from the transformed data for each participant 6_Scripts     - Unity Project         contains the complete project of our user-controlled experiment (Unity 2018.4.1f1, HTC VIVE Pro Eye HMD)     - Data Processed         contains scripts that undertake the pre-processing steps for converting the raw data to the transformed/frame data in the transformed and frame folders.         conatins scripts for continuous annotation, behavior and physiological data analysis and visualization.     - CEAP-360VR_Baseline         contains scripts to generate processed behavioral and physiological data with V-A labels for deep learning experiments and features for machine learning experiments.         contains scripts to run ML and DL experiments under both  subject-dependent and subject-independent model. ## Dataset Description The CEAP-360VR Dataset [Description.pdf](https://github.com/cwi-dis/CEAP-360VR-Dataset/blob/master/CEAP-Dataset%20Description.pdf) introduces the dataset description and key steps in the stage of data acquisition and pre-processing. ## Dataset License CEAP-360VR dataset is licensed under a [Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license]. ## Citation Please cite our paper in any published work that uses this dataset as follows: - Plain Text T. Xue, A. El Ali, T. Zhang, G. Ding, and P. Cesar, "CEAP-360VR: A Continuous Physiological and Behavioral Emotion Annotation Dataset for 360° Videos," in IEEE Transactions on Multimedia, doi: 10.1109/TMM.2021.3124080. - BibTex @ARTICLE{Xue2021CEAP-360VR,   author={Xue, Tong and Ali, Abdallah El and Zhang, Tianyi and Ding, Gangyi and Cesar, Pablo},   journal={IEEE Transactions on Multimedia},    title={CEAP-360VR: A Continuous Physiological and Behavioral Emotion Annotation Dataset for 360° Videos},    year={2021},   volume={},   number={},   pages={1-1},   doi={https://doi.org/10.1109/TMM.2021.3124080}} ## Usage  1. We have performed the time alignment of different types of data and      videos for each participant, as well as the proceesing scripts that      can be used to generate both the transformed and frame data.        Researchers can run their analysis methods on them.          2. For researchers who want to try other data processing methods, you can directly use the raw data. ## About  The CEAP-360VR Dataset is maintained by Key Laboratory of Digital Performance and Simulation Technology at Beijing Institute of Technology and the Distributed & Interactive Systems (DIS) research group at Centrum Wiskunde & Informatica . Contact the authors - Tong Xue: xuetong@bit.edu.cn, xue.tong@cwi.nl - Abdallah El Ali: abdallah.el.ali@cwi.nl

opencc-by-4.0Jun 2024View details →

Drone videos and their annotations of passing sheep (for counting purpose)

The data below are part of the European H2020 project ICAERUS regarding the livestock monitoring use case. More information here : https://icaerus.eu/ Counting sheep is particularly challenging for farmers with hundreds of animals in a flock. Our objective is to develop methodology based on computer vision to count sheep when they are passing in a corridor to coming back in a park or a pen. This first dataset will support our work.  The dataset encompasses 4 .MP4 videos from drone (DJI mavic 3 Enterprise or Thermal) of around 50 sheep crossing a gate.  The videos were taken from 5m to 10m of height and to an horizontal distance of the gate from 0m to 10m.  The videos come from previous datasets (https://doi.org/10.5281/zenodo.10400302) but have been modified to facilitate annotation: the duration of each video has been reduced to 30 seconds or less with 9 frames per second. Image size has been modified to 1440x1080 pixels. Annotation files are available in MOTS and YOLO formats. This dataset encompasses the following data: -----Videos: a directory were the videos are stored (4 videos, RGB images taken from 5 or 10m of altitude; images size of 1440x1080, videos taken with DJI MAVIC3T). -------------crop_23.11.23-XX: a directory by flight containing the video of a unique flight, with the date (YY.MM.DD) and XX representing a mission number -----MOT1.1: a directory with the annotations of cows at the MOTS format -------------MOT1.1_crop_23.11.23-XX: a directory by flight containing the annotations, one annotation file referred to an unique image and have the same name except the extension -----YOLO1.1: a directory with the annotations of cows at the YOLO format -------------YOLO1.1_crop_23.11.23-XX: a directory by flight containing the annotations, one annotation file referred to an unique image and have the same name except the extension More videos will be published in the next months. For more information, please contact: adrien.lebreton@idele.fr  The authors are opened to any collaborations on this topic.

opencc-by-4.0Sep 2024View details →

Grape Bunch Video Dataset with Berry Annotations and Tracking Data

The dataset consists of videos of Bonarda grape bunches in a mature state. A total of 100 grape bunches were collected from five consecutive rows of the same plot at the "Finca de Sancho" located in Lavalle, Mendoza, Argentina, on March 21, 2023. Each bunch was assigned a unique identifier after collection. For video capture, two setups were built, referred to as setup 1 and setup 2. Each consisted of a stand to hold a bunch against a smooth white background. This background was carefully chosen to facilitate precise segmentation of the grapes in the images, ensuring more efficient and accurate object detection. Additionally, a curved structure with a 10x7 grid of QR codes was placed behind the bunch. These codes were not used in this dataset but were included for potential future studies. The setups were mounted outdoors, taking advantage of natural lighting to obtain a more realistic representation of the grape bunches. Video capture sessions were conducted over three consecutive days, lasting between 4 to 7 hours each day. The first session took place on the same day the bunches were harvested. Throughout the sessions, natural light fluctuated due to intermittent cloud cover, introducing variations in ambient lighting. Video recording was performed by two individuals referred to as capturer A and capturer B. The recording devices were the cameras of two smartphones, a Samsung Galaxy S20 FE and a Motorola G200. The Open Camera application (<a href="https://opencamera.org.uk/" target="_new" rel="noopener">https://opencamera.org.uk/</a>) was used, configured to automatically capture 5-second videos at a resolution of 720 pixels wide by 1280 pixels high (portrait orientation), at 30 frames per second. A fixed focal distance of 30 cm was maintained, and the white balance was manually adjusted according to the natural light variations to maintain good image quality.  Three camera movements were defined for video capture: two systematic movements named "horizontal 180º" and "vertical 180º", and a third movement called "freestyle." The horizontal and vertical movements involved sweeping the camera from left to right and from bottom to top, respectively, while keeping the bunch centered in the frame and at a distance of approximately 30 cm. The freestyle movement consisted of random movements, maintaining the bunch centered as best as possible within the frame. The capture protocol was as follows: First, a grape bunch was placed in each setup. Capturer A at setup 1 captured five videos for each of the three defined camera movements, while capturer B did the same at setup 2. Afterward, the capturers switched positions, with capturer B taking videos at setup 1 and capturer A at setup 2. The bunches were then replaced with new ones, and the process was repeated until all 100 bunches were captured. As a result, the dataset contains five videos for each of the three camera movements, captured by both individuals for each of the 100 bunches. This produces a dataset consisting of 3,000 videos: 1,000 videos for each camera movement and 30 videos per bunch.  It is worth noting that, due to the manual nature of the capture process, some variations occurred in the number of videos recorded per bunch. In some cases, 9 videos were recorded instead of 10, in one instance only 8 were captured, and in one case 11 videos were obtained. These minor discrepancies are mainly due to human error in counting the recorded videos, but they do not affect the overall quality or integrity of the dataset.   The dataset also includes berry annotations for each video, provided in JSON files. These files specify, for each frame of the video, the pixel coordinates of each berry's center and the radius it occupies in the image, also in pixels. The berry detections were obtained through inference using a deep learning architecture called CircleNet, which was specifically trained for this dataset. In addition, the dataset contains berry tracking data, provided in CSV files. These files indicate which berries are the same across different frames of the video, through a unique berry identifier. The tracking was generated using a custom algorithm developed specifically to produce these tracks.

opencc-by-4.0Aug 2020View details →

Annotated educational videos and subtitles (EDUVSUM)

We have collected educational videos with subtitles from three popular e-learning platforms: Edx,YouTube, and TIB AV-Portal that cover the following topics: crash course on history of science and engineering, computer science, python and web programming, machine learning and computer vision, Internet of things (IoT), and software engineering. In total, the current version of the dataset contains 98 videos with ground truth values annotated by a user with an academic background in computer science.     Cite our work in order to use the resources (<a href="https://arxiv.org/abs/2010.13626">https://arxiv.org/abs/2010.13626</a> ) <a href="https://github.com/VideoAnalysis/EDUVSUM">https://github.com/VideoAnalysis/EDUVSUM</a> <a href="https://github.com/VideoAnalysis/VideoAnnotationTool">https://github.com/VideoAnalysis/VideoAnnotationTool</a>   <pre><code>@article{ghauri2020eduvsum, title={Classification of Important Segments in Educational Videos using Multimodal Features}, author={Ghauri, Junaid Ahmed and Hakimov, Sherzod and Ewerth, Ralph}, Conference={International Workshop on Investigating Learning During Web Search (IWILDS 2020) co-located with CIKM}, year={2020} }</code></pre>

opencc-by-4.0Aug 2024View details →

Video annotations from FK210506

Video annoations from 4 ROV dives as part of FK210605

opencc-by-4.0Jul 2022View details →

YM2413-MDB: A Multi-Instrumental FM Video Game Music Dataset with Emotion Annotations

YM2413-MDB is an 80s FM video game music dataset with multi-label emotion annotations. It includes 669 audio and MIDI files of music from Sega and MSX PC games in the 80s using YM2413, a programmable sound generator based on FM. The collected game music is arranged with a subset of 15 monophonic instruments and one drum instrument. They were converted from binary commands of the YM2413 sound chip. Each song was labeled with 19 emotion tags by two annotators and validated by three verifiers to obtain refined tags For more detailed information about the dataset, please refer to our paper: <a href="https://arxiv.org/abs/2211.07131">YM2413-MDB: A Multi-Instrumental FM Video Game Music Dataset with Emotion Annotations</a>. File Description 1) Pure data - original_vgms: crawled vgm files from <a href="https://www.smspower.org/">SMS POWER</a> and <a href="https://vgmrips.net/packs/">VGMRIPs</a> - wav: rendered vgm files using <a href="https://github.com/vgmrips/vgmplay">VGMPlay</a>   2) MIDI data - midi/vgmplay_log_to_midi: converted midi files - midi/adjust_tempo: add postprocessing(metrically aligned using wav_downbeat files) after midi conversion - midi/adjust_tempo_remove_delayed_inst: add postprocessing(metrically aligned using wav_downbeat files, remove delayed instrument) after midi conversion   3) Metadata - emotion_annotation/verified_annotation.csv: contains emotion annotation for each songs - tags_kor_eng.txt: Korean <-> English tag dictionary   4) Useful middle-time step data - wav_downbeat: extracted downbeat values using TCNBeatTracker of <a href="https://github.com/CPJKU/madmom">madmom</a> - vgm_txts: disassembled vgm files as txt using <a href="https://github.com/vgmrips/vgmtools#vgm-text-writer-vgm2txt">vgm2txt</a> - ydr: YM2413 Disassembly Raw(YDR). command list of vgm files. generated by reading vgm_txts   Update Log - version 1.0.1: Fix ticks per beat value adjust to tempo where tempo values are not 150. Also, madmom downbeat files are updated from DBNBeatTracker(ISMIR, 2015) to TCNBeatTracker(Newer one EUSIPCO, 2019). - version 1.0.2: <a href="https://github.com/jech2/YM2413-MDB/issues/2">Wrong emotion tag issue in the verification annotation file was fixed.</a>