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.
3,905
datasets available to search
ShareScore release 0.9.0
Dataset results
3,905 results for “eye”
Resting state with closed eyes for patients with depression and healthy participants
Open the record for dataset details and reuse information.
Two sessions of resting state with closed eyes for patients with depression in treatment course (NFB, CBT or No treatment groups)
Open the record for dataset details and reuse information.
PsPM-AOB: Eye tracker (including pupillometry) measurements from auditory oddball tasks
<p>This dataset includes eye tracker (including pupillometry) measurements from auditory oddball tasks with ITIs of 1, 2, and 3 s (in groups 1, 2, and 3 respectively). Also included are task information, keypress responses, keypress response times and key correctness for each of 66 healthy unmedicated participants (40 females and 26 males aged 24.2+/-3.9 years) participating in auditory oddball tasks. Stimuli consist of sine tones (50-ms length; 10-ms ramp; 440 or 660 Hz).</p>
PsPM-EWO: Eye tracker (including pupillometry) measurements from emotional-words tasks
<p>This dataset includes eye tracker (including pupillometry) measurements for 37 healthy unmedicated participants (25 females and 12 males, age range: 18 - 34 years, mean age: 26.2 +/- 4.7 years) participating in an emotional-words task. In each session, participants were presented with 50 neutral and 50 negative five-letter nouns from the Berlin Affective Word List Reloaded (Võ et al., 2009). Also included are task information, keypress responses, keypress response times.</p>
PsPM-RRM1-2: SCR, ECG, respiration and eye tracker measurements in response to electric stimulation or visual targets
<p>This dataset includes skin conductance response (SCR), electrocardiogram (ECG), respiration and eye tracker (including pupillometry) measurements for each of 29 healthy unmedicated participants (7 males and 22 females aged 23.5 +/- 3.6 years) in response to 10 discomforting electric stimulations to the forearm (RRM1) or 10 visual targets in a visual detection task (RRM2). The sample partly overlaps with data set <a href="https://doi.org/10.5281/zenodo.1292568">PsPM-FR</a>. Some participants did not take part in RRM1 or RRM2 such that there are 25 recordings for RRM1 and 26 recordings for RRM2. Electric shock stimuli are 0.2 ms wide square current pulse repeated at 500 Hz for 500 ms and individually adjusted amplitude just below the pain threshold. Visual stimuli are red crosses (+) embedded in a white digit stream; each stimulus is presented during 200 ms and separated by a 800 ms blank interval. ITI is selected randomly on each trial from 40 s, 45 s or 50 s. A baseline period with distractors but no targets concludes experiment RRM2. (This is in contrast to the methods description in Bach et al. (2016), according to which the baseline period was randomly either in the beginning or at the end of the experiment. This discrepancy was caused by an error in the code that controlled the experiment presentation.)</p>
MAMEM Phase I Dataset - A dataset for multimodal human-computer interaction using biosignals and eye tracking information
<p>This dataset combines multimodal biosignals and eye tracking information gathered under a human-computer interaction framework. The dataset was developed in the vein of the MAMEM project that aims to endow people with motor disabilities with the ability to edit and author multimedia content through mental commands and gaze activity. The dataset includes EEG, eye-tracking, and physiological (GSR and Heart rate) signals along with demographic, clinical and behavioral data collected from 36 individuals (18 able-bodied and 18 motor-impaired). Data were collected during the interaction with specifically designed interface for web browsing and multimedia content manipulation and during imaginary movement tasks. Alongside these data we also include evaluation reports both from the subjects and the experimenters as far as the experimental procedure and collected dataset are concerned. We believe that the presented dataset will contribute towards the development and evaluation of modern human-computer interaction systems that would foster the integration of people with severe motor impairments back into society.</p>
THÖR - eye-tracking
<p><strong>THÖR</strong> is a dataset with human motion trajectory and eye gaze data collected in an indoor environment with accurate ground truth for the position, head orientation, gaze direction, social grouping and goals. THÖR contains sensor data collected by a 3D lidar sensor and involves a mobile robot navigating the space. In comparison to other, our dataset has a larger variety in human motion behaviour, is less noisy, and contains annotations at higher frequencies.</p> <p><strong>THÖR eye-tracking -</strong> data of the participant (Helmet number 9) from the Tobii Glasses included in this dataset. The entire data from the experiment was recorded into two recordings - "Recording011" and "Recording012".</p> <p>The folder "RawData" consists of exported data from Tobii Pro Lab software using a filter called "Tobii-IVT Attention filter" (velocity threshold parameter set to 100 degrees/second), which is a recommended method for dynamic situations. The recommendation was given by the equipment manufactures Tobii Pro and from other researchers. For further information, please refer to https://www.tobiipro.com/siteassets/tobii-pro/user-manuals/Tobii-Pro-Lab-User-Manual/?v=1.86 (Appendix-B, page 85).</p> <p>The recording start times are as follows:<br> Tobii recording011 starttime: 13:34:37.267<br> Tobii recording012 starttime: 14:36:17.730</p> <p>1. "Synchronized_Qualisys_Tobii.mat" file, which consists of all the synchronised data between Qualisys data and Tobii eye-tracker data using timestamps matching. The columns (headers) in this mat file respectively represent - 'timestamp', 'Pos_X', 'Pos_Y', 'Pos_Z', 'Head_R', 'Head_P', 'Head_Y', 'GazepointX', 'GazepointY', 'Gazepoint3D_X', 'Gazepoint3D_Y', 'Gazepoint3D_Z', 'Gazedirectionleft_X', 'Gazedirectionleft_Y', 'Gazedirectionleft_Z', 'Gazedirectionright_X', 'Gazedirectionright_Y', 'Gazedirectionright_Z', 'Pupilpositionleft_X', 'Pupilpositionleft_Y', 'Pupilpositionleft_Z', 'Pupilpositionright_X', 'Pupilpositionright_Y', 'Pupilpositionright_Z', 'Pupildiameterleft', 'Pupildiameterright', 'Gazeeventduration', 'Eyemovementtype_index', 'Fixationpoint_X', 'Fixationpoint_Y', 'Gyro_X', 'Gyro_Y', 'Gyro_Z', 'Accelerometer_X', 'Accelerometer_Y', 'Accelerometer_Z'.</p> <p>2. A matlab script "Synchronizing_Qualisys_Tobii.m" used for matching the timestamps of Qualisys and Tobii and generate a data file "Synchronized_Qualisys_Tobii.mat" that can be found it git repository.</p> <p>3. All the associated data required for running the matlab script.</p> <p>Please note that 'Timestamp', 'Pos_X', 'Pos_Y', 'Pos_Z', 'Head_R', 'Head_P', 'Head_Y','GazepointX', 'GazepointY' represent the timestamps (matched to Tobii timestamps using nearest neighbor search), position and head orientation from Qualisys data and rest of the data is from Tobii eye-tracker. For more information regarding the eye-tracker data, please refer to https://www.tobiipro.com/siteassets/tobii-pro/user-manuals/Tobii-Pro-Lab-User-Manual/?v=1.86 (Section 8.7.2.1, page 68).</p> <p> </p>
The Contributionsof Eye Gaze Fixations and Target-Lure Similarity to Behavioral and fMRI Indices of Pattern Separation and Pattern Completion
Open the record for dataset details and reuse information.
More precise tracking of horizontal than vertical target motion with both the eyes and hand
<p>Those files contain individual data from a large cohort of participant (N=62). </p> <p>In the excel file (DATAmain), each sheet presents one set of variables (with individual value for each trial).</p> <p>This file contains information regarding eye and hand tracking performance (distance+lags), as well as smooth pursuit gains. </p> <p>The other files contain data that we used for the detailed analysis of saccades and lags, as well as the scripts that can be run with Perl. One script is for analysing the lag (Danion.pl) and the other one for analysing the saccades (saccades.pl). The other files (.txt and .dat) that were used for these analyses. Note that some library is needed (common_subroutines, draw_figure, and for the anova’s routines_that_use_R), meaning that you need to have R installed. </p> <pre>Regarding data acquisition we employed a program called Docometre that can be uploaded at the following address: http://139.124.68.1/buloup/index.php?selectedMenu=DOCoMETRe&lang=_fr When this program is installed, it needs to be run with BaselineTracking.dcm We also provide .BAS and .T91 files that correspond to the compiled version of each pattern Regarding visual stimuli, another program called ICE needs to be installed on a separate computer that receives information (target+cursor) from docometer, it can be uploaded at : https://trello.com/b/EtNCNrZH/icehttps://trello.com/b/EtNCNrZH/ice ICE needs to be run with Visuomotor.ice Visuomotor.icepro Visuomotor.txt and Visuomotor.icemat in the respective folder (icepro in Protocol folder, icemat and ice in Scenario Folder, and txt in Serie folder) Note that both Docometre and ICE need to be run with similar equipement as our (including Adwin Gold systems, Megatron joystick, video screen, graphic cards, and desktop eyelink providing analog signals to docometre). Adequate numbering of analogic channels needs also to be ensured. </pre>
Anthropogenic emissions of CH4, N2O, F-gases and BC from GAINS, for EU-countries plus CH, NO, UK developed under the EYE-CLIMA project - March 2025 update
<p><span>As part of the EYE-CLIMA project, GAINS emission data for CH<sub>4</sub>, N<sub>2</sub>O, BC and selected F-gases (HFC-125, HFC-134a, HFC-143a, HFC-23, HFC-32 and SF<sub>6</sub></span>) were released for all EU-27 countries plus UK, Switzerland, and Norway for the period 1990 to 2020 (with exception of F-gases, from 2005 only, and BC/CH<sub>4</sub> emissions from agricultural waste burning, from 2000). Results have been documented in EYE-CLIMA deliverable D2.8 (<a href="http://folk.nilu.no/~rthompson/eyeclima_reports/EYECLIMA_D2.8.pdf">http://folk.nilu.no/~rthompson/eyeclima_reports/EYECLIMA_D2.8.pdf</a>), and they are publicly available at the Zenodo repository under <a href="https://doi.org/10.5281/zenodo.11032177">https://doi.org/10.5281/zenodo.11032177</a>. All data is available on a 0.1°x0.1° grid and in monthly resolution. Emissions are attributed to the respective source categories according to GNFR.</p> <p>The motivation of an update resulted from the need to extending the emission data time series to 2023. With underlying statistics and national emission data currently available till 2022 only (the latter submitted to UNFCCC only by December 2024), the historical data series also could only be established for 2022. Here we use the GAINS scenario feature to extrapolate between 2022 historical data and the first scenario point, 2025 which is based on IEA’s Word Energy Outlook 2023 (https://www.iea.org/reports/world-energy-outlook-2023). Obviously, this also means that emission results for 2023 are not any more based on robust statistics but represent an extrapolation.</p> <p>Extrapolation of spatially explicit data is only possible when the spatial resolution conveys a realistic signal. For the sector “agricultural waste burning” (files with “AWB” as sector, see notation below) spatial allocation is based on actual observation from satellites. As such data products on agricultural fires have been made available until 2022 only, no spatial or temporal signal exists for 2023. The time series provided thus has to end in 2022. No recommendation can be given to modellers, other than to either use 2022 also for 2023 (understanding that the pattern will be strikingly different) or to use a five-year average (which will remove a lot of spatial specificity).</p> <p>The updated dataset covers files as follows (internally, all files now carry version number V05):</p> <p>ALL_FLUX_ALL_EUR_MOD_MONTH_19900101_20231231_GAINS_IIASA_V05.csv</p> <p>BC_FLUX_ALL_EUR_MOD_MONTH_19900101_20231231_GAINS_IIASA_V05.nc</p> <p>BC_FLUX_AWB_EUR_MOD_MONTH_20000101_20221231_GAINS_IIASA_V05.nc</p> <p>CH4_FLUX_ALL_EUR_MOD_MONTH_19900101_20231231_GAINS_IIASA_V05.nc</p> <p>CH4_FLUX_AWB_EUR_MOD_MONTH_20000101_20221231_GAINS_IIASA_V05.nc</p> <p>HFC_FLUX_ALL_EUR_MOD_YEAR_20050101_20231231_GAINS_IIASA_V05.nc</p> <p>N2O_FLUX_ALL_EUR_MOD_MONTH_19900101_20231231_GAINS_IIASA_V05.nc</p> <p>SF6_FLUX_ALL_EUR_MOD_YEAR_20050101_20231231_GAINS_IIASA_V05.nc</p> <p>This is version 2.0 of the dataset. It extends from version 1.0 by covering into the year 2023, but also benefits from a number of additional GAINS improvements. Emissions of emitted compounds are provided as kg/m²/s. File names follow the notation developed for the H-Europe project EYE-CLIMA, i.e. species _ variable-type _ sector _ region _ method (MOD=model) _ timestep _ fromTime _ toTime _ model _ institute _ version . filetype.</p> <p>This version is available at <a href="https://doi.org/10.5281/zenodo.15536170">https://doi.org/10.5281/zenodo.15536170</a>. The generic address of the dataset is <a href="https://doi.org/10.5281/zenodo.10886780">https://doi.org/10.5281/zenodo.10886780</a>, resolving to the latest update available at Zenodo. No further updates are planned in EYE-CLIMA, so this version is expected to also reflect the final update within the project.</p> <p>Compared to version 1.0, GAINS benefitted from a number of new developments such as the following:</p> <p>*) Previously, GAINS has been available in five-year timesteps only (with the aim of allowing for scenarios at that resolution). For data version 1.0, a makeshift solution was found to convert into annual data. A recent update now allows, for historic data, to store and retrieve information on an annual basis (from 1990).</p> <p>*) The energy data were obtained from IEA’s world energy balances 2024 (July version, https://www.iea.org/data-and-statistics/data-product/world-energy-balances#documentation), extending into 2022 and extrapolated towards 2025, downscaled from IEA to GAINS sectors and sub-sectors. Additionally, the annual activity of industrial production is estimated using a linear approach, based on five-year timestep data.</p> <p>*) Agricultural statistics were retrieved from Eurostat (and from FAO globally) and extended to 2022, extrapolated towards 2025.</p> <p>*) Interpretation of GAINS data was reconfirmed and updated in consultations with national experts of multiple EU countries. While the process resulted in revised emission projections to be used in the Clean Air Outlook 4 (see <a title="Protected by Check Point: https://environment.ec.europa.eu/topics/air/clean-air-outlook_en" href="https://protect.checkpoint.com/v2/r02/___https:/environment.ec.europa.eu/topics/air/clean-air-outlook_en___.YzJlOmlpYXNhOmM6bzoyYzdiNDRhNDI4Njc3ZjI5MGFjMTU1N2I2OWVmNzM2ZTo3OjE5OTM6ZTFiY2IzMDMxZGViNGE0MjI0ODRmNWQ4NzA3ZDY3Njc4M2U2NzUxNmEwNzQ0ODViNDBhODc1NmNhZmMzY2FlMjpoOkY6Tg"><span lang="EN-GB">https://environment.ec.europa.eu/topics/air/clean-air-outlook_en</span></a><span lang="EN-GB">). While the details of improvements on the individual aspects cannot be disclosed, they are useful to describe historic data most adequately, and have been integrated also in this assessment. That not only leads to changes in absolute emissions for a given year, but also affects trends that now are more plausible and confirmed through the exchange with the national experts.</span></p> <p><span lang="EN-GB">*) Technical adjustments have improved the precision of temporal allocation of emissions and the conversion of grid sizes to actual area.</span></p>
Closed and Open Eyes EEG Data
<p>Seven volunteers agreed to participate in the study. Their mean age was 29.67 years (range 24 – 56 years). The participants indicated that they did not have hearing or visual impairments. </p><p>Gold cup electrodes were O1 and O2 placed following the 10-20 International System for electrode placement and attached to the subject scalp using a conductive paste. Electrode-skin impedances were checked to be below 15 kΩ at all electrodes. The reference and ground electrodes were placed in the Fp2 and A2 positions, respectively, where the absence of hair facilitates their placement, thus optimizing the setup time and EEG signal quality. </p><p>The signals were captured using the hardware presented in [1], and the signal processing algorithms were executed on a PC using Matlab, allowing us to repeat the simulations offline with different parameters.</p><p>During the experimental sessions, the signals from the two channels were recorded for a total duration of 10 minutes per participant. Specifically, the recording process involved 60 seconds of signal acquisition while the participant had their eyes open, followed by another 60 seconds of signal acquisition while the participant had their eyes closed. To indicate the transition between the two eye states, a sound alert was played for the participant. Once the electrodes had been placed and the impedance checked to be below 15 kΩ, the recordings started without individual calibration for any of the participants. All the experiments were conducted in a sound-attenuated and controlled environment. Participants were seated in a comfortable chair and asked to be relaxed and focused on the task, trying to avoid any distractions or external stimuli. To mimic real-life conditions, the participants were allowed to freely move their gaze during the eye-open tasks, without the requirement of maintaining fixation on a specific point. To reduce possible artifacts, participants were asked not to move or speak during the experiments. After each recording session, data for each subject were visually inspected and the recording was repeated if any of them was corrupted by a high level of noise or artifacts.</p><p>Data is organized in a folder for each subject (S1, S2, S3, etc.). Inside each subject folder, another folder named 'PP' contains the EEG recordings in a .csv file. </p><p>Each .csv file contains 3 columns: timestamp, O1 channel and O2 channel.</p><p> </p><p>[1] Laport F, Dapena A, Castro PM, Iglesias DI, Vazquez-Araujo FJ. Eye State Detection Using Frequency Features from 1 or 2-Channel EEG. Int J Neural Syst. 2023 Dec;33(12):2350062. doi: 10.1142/S0129065723500624. Epub 2023 Oct 12. PMID: 37822240</p>
Eye tracking videos and raw data of breathing recognition attempts in simulated out-of-hospital cardiac arrest
<div> <div> <div> <p>This dataset comprises eye tracking videos and raw data documenting attempts to recognize breathing in simulated out-of-hospital cardiac arrest scenarios.</p> <p>The data were recorded using an Ergoneers Dikablis head-mounted eye tracker.</p> <p>Our analysis of this data resulted in the publication of two studies: Study 1, available at <a href="https://doi.org/10.1097/SIH.0000000000000617" target="_blank" rel="noopener">https://doi.org/10.1097/SIH.0000000000000617</a>, and Study 2, accessible at <a href="https://doi.org/10.25894/ijfae.2307" target="_blank" rel="noopener">https://doi.org/10.25894/ijfae.2307</a></p> <p> </p> <p>Version 2 is up-to-date.</p> <p>In Version 1:</p> <ul> <li>the doi for Study 2 was incorrect</li> <li>data for participant #51 of Study 1 were missing</li> </ul> </div> </div> </div>
Healthy Eyes, Happy Child
<p>Healthy Eyes, Happy Child” (HEHC) projetc investigated the impact of spectacle correction on the well-being of children from ages 6 to 12 years in rural primary schools in the Pinetown Education District of Kwa-Zulu Natal, South Africa.</p>
A Real-Time Eye-Tracking Dataset for Autism Severity Classification Using Deep Learning
<p>Eye-Tracking (ET) technologies have shown significant potential in autism research, providing critical insights into gaze patterns and their correlation with autism severity. However, a persistent challenge in developing Deep Learning (DL) models for ET analysis is the lack of publicly available, annotated datasets tailored for specific tasks. In order to close this gap, we present a novel, meticulously annotated resource designed to classify autism severity based on ET data. This dataset consists of 4,000 high-resolution (416×416 pixels) eye images derived from video recordings of 40 participants, evenly distributed across four autism severity groups: low, mild, medium, and high.</p> <p>Each participant's video was processed to extract 50 frames per session, capturing diverse gaze behaviors such as fixations, saccades, and smooth pursuits. Both left and right eye images were segmented from these frames, yielding 100 images per participant and ensuring balanced representation across severity categories (1,000 images per group). The dataset is annotated with detailed metadata, including subject ID, frame number, autism severity level, and eye type (left or right), providing a robust foundation for precise feature extraction and analysis.</p> <p><span>Facilitating its application in DL model development, this dataset addresses a critical gap in the limited availability of ET datasets. It provides a robust benchmark for autism severity classification, establishing a foundational resource for advancing Machine Learning(ML) research in the domain of autism</span><span>. This dataset serves as a critical resource for advancing ET-based classification models, fostering accurate and efficient assessment of autism severity, and supporting broader autism research.</span></p>
Data to Three-Dimensional Binocular Eye-Hand Coordination in Normal Vision and with Simulated Visual Impairment
<p>This record contains experimental and analysis scripts (written in Matlab) as well as raw and processed data to reproduce the results shown in:</p> <p>Maiello, G., Kwon, M. & Bex, P.J. (2018) Three-dimensional binocular eye--hand coordination in normal vision and with simulated visual impairment. <em>Experimental Brain Research</em>. https://doi.org/10.1007/s00221-017-5160-8</p>
The INI-30 Dataset : Event Camera for Eye Tracking
<p>The Ini-30 dataset is collected with two event cameras mounted on a glass frame. Each DVXplorer sensor (640 × 480 pixels) is attached on the side of the frame. The power supply was provided via a 2 meter cable connected from the cameras to a computer, which provided enough freedom of movement. Differently from [2, 24], the participants were not instructed to follow a dot on a screen, but rather encouraged to look around to collect natural eye movements. As shown in Fig. 1, the event cameras were securely screwed on a 3D-printed case attached to the side of the glass frame. The data was annotated based on accumulated linearly decayed events by defining the pixel intensity as function of the linear accumulation of previous pixel intensity. Next we labeled the position of the pupil in the DVS’s array manually, using an assistive labeling tool. We discarded the first 20ms of events to ensure the eye was visible and annotations met the level of image-based annotators. The number of labels per recording was intentionally variable, spanning from 475 to 1’848 with a time per label ranging from 20.0 to 235.77 milliseconds depending on the overall duration of the sample. This setup allows for unconstrained head movements, enables to capture event data from eye movement in a ”in-the-wild” setting and allows the generation of a representative, unique, diverse and challenging dataset.<br><br>NOTE : the annotations relates to the ellipse of the pupil on the image</p>
Viewing behavior and vertical eye-level light for non-image-forming effects
<p>When considering non-image-forming (NIF) light effects on people, knowing the light vertically at eye-level is necessary. However, people are dynamic in their behavior and constantly change their viewing direction. This means that light measured vertically towards a constant direction might differ from the actual light that reaches people’s eyes. If the difference is large, viewing behavior might need to be included in lighting design measurements and simulations predicting the potential of the light to induce NIF light effects. This dataset was collected during an experiment on the difference between the actual dynamic eye-level light of office workers while seated at a desk (dynamic condition) and light measured statically towards a computer screen (static condition). The dataset was collected to test the hypothesis: "There is a significant and relevant difference between simultaneously measured static and dynamic light conditions in an office environment occupied by one user." It includes measured and simulated light quantities (illuminance, alpha-opic quantities according to CIE S026 and light-driven alertness according to the non-visual direct response model) together with participants' measured face orientation (horizontal and vertical) in an office environment with a single user.</p>
Data on eye movements in people with glaucoma and peers with normal vision
<p>Eye movements were recorded from 44 elderly glaucoma patients and 32 age-similar healthy vision controls whilst watching three separate small video clips.</p>
Data to "Retinal Blur from Natural Scenes and Eye Shape"
<p>This record contains experimental and analysis scripts (written in Matlab) as well as raw and processed data to reproduce the results shown in:</p> <p><strong>Maiello, G</strong>., Harrison, W. J., Vera-Diaz, F. A., & Bex, P. J. (in preparation). Retinal Blur from Natural Scenes and Eye Shape.</p>
Atari-HEAD: Atari Human Eye-Tracking and Demonstration Dataset
<p>Version 4 of the dataset is available (Sep 19 2019)!</p> <p>Note this version has significantly more data than Version 2. </p> <p>Dataset description paper (full version) is available!</p> <p>https://arxiv.org/pdf/1903.06754.pdf (updated Sep 7 2019)</p> <p>Tools for visualizing the data is available!</p> <p>https://github.com/corgiTrax/Gaze-Data-Processor</p> <p> </p> <p><strong>=========================== Dataset Description ===========================</strong></p> <p>We provide a large-scale, high-quality dataset of human actions with simultaneously recorded eye movements while humans play Atari video games. The dataset consists of 117 hours of gameplay data from a diverse set of 20 games, with 8 million action demonstrations and 328 million gaze samples. We introduce a novel form of gameplay, in which the human plays in a semi-frame-by-frame manner. This leads to near-optimal game decisions and game scores that are comparable or better than known human records. For every game frame, its corresponding image frame, the human keystroke action, the reaction time to make that action, the gaze positions, and immediate reward returned by the environment were recorded.</p> <p> </p> <p>Q & A: Why frame-by-frame game mode?</p> <p><strong>Resolving state-action mismatch</strong>: Closed-loop human visuomotor reaction time is around 250-300 milliseconds. Therefore, during gameplay, state (image) and action that are simultaneously recorded at time step t could be mismatched. Action at time t could be intended for a state 250-300ms ago. This effect causes a serious issue for supervised learning algorithms, since label at and input st are no longer matched. Frame-by-frame game play ensures states and actions are matched at every timestep.</p> <p><strong>Maximizing human performance</strong>: Frame-by-frame mode makes gameplay more relaxing and reduces fatigue, which could normally result in blinking and would corrupt eye-tracking data. More importantly, this design reduces sub-optimal decisions caused by inattentive blindness.</p> <p><strong>Highlighting critical states that require multiple eye movements</strong>: Human decision time and all eye movements were recorded at every frame. The states that could lead to a large reward or penalty, or the ones that require sophisticated planning, will take longer and require multiple eye movements for the player to make a decision. Stopping gameplay means that the observer can use eye-movements to resolve complex situations. This is important because if the algorithm is going to learn from eye-movements it must contain all “relevant” eye-movements.</p> <p> </p> <p><strong>============================ Readme ============================</strong></p> <p>1. meta_data.csv: meta data for the dataset., including:</p> <ul> <li> <p>GameName: String. Game name. e.g., “alien” indicates the trial is collected for game Alien (15 min time limit). “alien_highscore” is the trajectory collected from the best player’s highest score (2 hour limit). See dataset description paper for details.</p> </li> </ul> <ul> <li> <p>trial_id: Integer. One can use this number to locate the associated .tar.bz2 file and label file.</p> </li> <li> <p>subject_id: Char. Human subject identifiers.</p> </li> <li> <p>load_trial: Integer. 0 indicates that the game starts from scratch. If this field is non-zero, it means that the current trial continues from a saved trial. The number indicates the trial number to look for.</p> </li> <li> <p>highest_score: Integer. The highest game score obtained from this trial.</p> </li> <li> <p>total_frame: Number of image frames in the .tar.bz2 repository.</p> </li> <li> <p>total_game_play_time: Integer. game time in ms. </p> </li> <li> <p>total_episode: Integer. number of episodes in the current trial. An episode terminates when all lives are consumed.</p> </li> <li> <p>avg_error: Float. Average eye-tracking validation error at the end of each trial in visual degree (1 visual degree = 1.44 cm in our experiment). See our paper for the calibration/validation process.</p> </li> <li> <p>max_error: Float. Max eye-tracking validation error. </p> </li> <li> <p>low_sample_rate: Percentage. Percentage of frames with less than 10 gaze samples. The most common reason for this is blinking.</p> </li> <li> <p>frame_averaging: Boolean. The game engine allows one to turn this on or off. When turning on (TRUE), two consecutive frames are averaged, this alleviates screen flickering in some games.</p> </li> <li> <p>fps: Integer. Frame per second when an action key is held down.</p> </li> </ul> <p> </p> <p>2. [game_name].zip files: these include data for each game, including:</p> <p>*.tar.bz2 files: contains game image frames. The filename indicates its trial number.</p> <p>*.txt files: label file for each trial, including:</p> <ul> <li> <p>frame_id: String. The ID of a frame, can be used to locate the corresponding image frame in .tar.bz2 file.</p> </li> <li> <p>episode_id: Integer (not available for some trials). Episode number, starting from 0 for each trial. A trial could contain a single trial or multiple trials.</p> </li> <li> <p>score: Integer (not available for some trials). Current game score for that frame.</p> </li> <li> <p>duration(ms): Integer. Time elapsed until the human player made a decision. </p> </li> <li> <p>unclipped_reward: Integer. Immediate reward returned by the game engine.</p> </li> <li> <p>action: Integer. See action_enums.txt for the mapping. This is consistent with the Arcade Learning Environment setup.</p> </li> <li> <p>gaze_positions: Null/A list of integers: x0,y0,x1,y1,...,xn,yn. Gaze positions for the current frame. Could be null if no gaze. (0,0) is the top-left corner. x: horizontal axis. y: vertical.</p> </li> </ul> <p> </p> <p>3. action_enums.txt: contains integer to action mapping defined by the Arcade Learning Environment. </p> <p> </p> <p><strong>============================ Citation ============================</strong></p> <p>If you use the Atari-HEAD in your research, we ask that you please cite the following:</p> <p>@misc{zhang2019atarihead,</p> <p> title={Atari-HEAD: Atari Human Eye-Tracking and Demonstration Dataset},</p> <p> author={Ruohan Zhang and Calen Walshe and Zhuode Liu and Lin Guan and Karl S. Muller and Jake A. Whritner and Luxin Zhang and Mary M. Hayhoe and Dana H. Ballard},</p> <p> year={2019},</p> <p> eprint={1903.06754},</p> <p> archivePrefix={arXiv},</p> <p> primaryClass={cs.LG}</p> <p>}</p> <p>Zhang, Ruohan, Zhuode Liu, Luxin Zhang, Jake A. Whritner, Karl S. Muller, Mary M. Hayhoe, and Dana H. Ballard. "AGIL: Learning attention from human for visuomotor tasks." In Proceedings of the European Conference on Computer Vision (ECCV), pp. 663-679. 2018.</p> <p>@inproceedings{zhang2018agil,</p> <p> title={AGIL: Learning attention from human for visuomotor tasks},</p> <p> author={Zhang, Ruohan and Liu, Zhuode and Zhang, Luxin and Whritner, Jake A and Muller, Karl S and Hayhoe, Mary M and Ballard, Dana H},</p> <p> booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},</p> <p> pages={663--679},</p> <p> year={2018}</p> <p>}</p> <p><br> <br> </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.