Joint infrared pupil images and near-corneal-plane spectral light exposure data under natural conditions across the adult lifespan – Dataset.

Detailed information on this dataset is provided in the corresponding data descriptor article. An excerpt from the "Data Records" section of that article is provided below. The dataset is licensed under CC BY 4.0.

The SpectroSense Dataset is organised in a structured folder system. Data were collected from 83 participants (first: SP001; last: SP113), with data from each participant stored in individual folders. In total, the dataset contains 29,664 recorded spectral irradiance and eye image pairs, along with 83 approximately three-minute-long calibration videos and further contextual data. Please note that the valid (corrected) timestamps of each sample are contained in the filenames of the (calibrated and uncalibrated) spectral irradiance samples and pupil images.

Additionally, four metadata files are included to support the use of the main data given in the SpectroSense Dataset.

0.1 demographic_data.csv contains the main demographic and ancillary data recorded before and during the trial for each participant (see Descriptive data in Methods). This includes the date of participation (date), start time of the health screening (begin), age (age) and age group (age_group) in years, sex assigned at birth (sex), handedness (handedness), visual aid status (visual_aid), body mass index (BMI) in kg/m², chronotype based on mid-sleep time (MSFsc), time since awakening in hours (time_awake), sleep duration in the night before the trial in hours (sleeping_hours), average weekly sleep duration in hours (SDweek), absolute sum of acutely reported caffeine consumption in mg (acute_sum), sum of acutely reported caffeine consumption relative to weight in in mg/kg (acute_sum_rel), absolute sum of habitually reported caffeine consumption in mg (habitual_sum), sum of habitually reported caffeine consumption relative to weight in in mg/kg (habitual_sum_rel), expertimenter-rated iris colour (iris_colour), weather during the field experiment phase (weather), KSS sleepiness rating (1–9, with 9 being most sleepy) in the indoor part of the trial (kss_pre), KSS sleepiness rating (1–9, with 9 being most sleepy) after the outdoor part of the trial (kss_post) and season during the participant in the experiment (season).

0.2 sat_files.csv contains a list of all saturated/erroneous spectrometer samples (after rapid dark-to-light changes, 410 samples in total). The saturated spectrometer samples are still included in the dataset as placeholders for completeness. For the calibrated and interpolated spectra, the samples take the spectral irradiance value “0” at all wavelengths and should be considered as missing and excluded when analysing the data.

0.3 sample_phase_tab.pdf shows a table summarising for every participant, which filenames have been taken under which experiment phase condition (“dark”, “laboratory”, “field” with transition samples being considered “mixed”). Please note that spectral irradiance measurements in the “dark” condition (taken in complete darkness) are affected by baseline noise as integration time was limited to 5 seconds, which can result in these spectral irradiance samples having values corresponding to up to ~1.5 lx (illuminance).

0.4 sum_tab.pdf shows a table that summarises for every participant, sex, age, the sequence of laboratory light conditions and the number of observations (pupil image and spectral irradiance pairs) in total and divided into experiment phase conditions (“dark”, “laboratory”, “field”, “mixed”).

The data in the "SpectroSense Dataset" (zip) folder are divided into participant-specific subdirectories. Each experimental session with a participant is stored in a participant-specific folder (SP001 to SP113), which contains the following subdirectories and files:

1. The subdirectory spectra contain the main data of this dataset, namely the spectral irradiance and pupil image samples collected at 10-second intervals.

1.1. Raw, uncalibrated spectroradiometer samples (format: YYYYMMDDHHMMSS.csv, e.g. 20210618104421.csv). The filenames contain the valid, corrected timestamps of each sample. These uncalibrated spectral light measurements include the following information:
1.1.1. Time: The time and date of the recording listed in the uncalibrated csv files are incorrect due to an issue with the internal time of the Raspberry Pi Computer used for sampling. However, as mentioned above, the corrected date and time are given in the filenames.
1.1.2.Serial Number, Integration time (microseconds), and Sensor temperature (°C), give contextual information on the spectroradiometer device. The latter two were used to correct for dark and thermal noise (see section “Dark and Thermal Noise Spectroradiometer Calibration” under “Technical Validation”).
1.1.3. Spectral irradiance measurements with irregular wavelength spacing (between 337 and 822 nm) are provided as Wavelengths (nm), Dark (counts), and Intensity (counts). These data are not interpretable without Ocean Insight-specific calibration. Spectral irradiance data from this dataset should, therefore be analysed using the calibrated and interpolated spectroradiometer samples (see below).

1.2. Calibrated and interpolated spectral irradiance samples (format: YYYYMMDDHHMMSS_calibrated_interp.csv, e.g. 20210618104421_calibrated_interp.csv): The csv files contain two columns: “wavelength” in nm given in the visible range (380–780 nm, 1 nm spacing) and “measurement” containing the corresponding irradiance given in W/m². Note that these spectra were resampled to 1 nm intervals between 380 and 780 nm, applying energy-preserving interpolation and linear interpolation for missing or defective pixels. The filenames contain the valid, corrected timestamps of each sample.

1.3. Pupil images: Still images of the participant's pupil (jpg format: 640 x 480 pixels, 96 dpi) taken with the “eye0” camera, corresponding to each spectroradiometer sample at 10-second intervals. The filenames contain the valid, corrected timestamps of each sample.

2. Pupil extraction settings screenshot. Each participant folder contains a PNG file (e.g. “SP001_PLsettings.png”) showing a screenshot of the pupil size extraction process in the Pupil Capture software. The screenshot provides metadata of the pupil extraction from the images, which was conducted to generate the processed data analysed in the registered report. This includes a visualisation of the 3D pupil model, including the pupil intensity range, which gives “the minimum darkness of a pixel to be considered as the pupil”. This pupil intensity range value took the value 10 at default and was adjusted (minimised) for every participant’s data so that the pupil was always fully covered while having as little leakage outside of the pupil as possible. Differences between participants occurred mainly due to differences in iris colour and differences in illumination of the eye during the field condition. The screenshot also contains “Pupil min” (10) and “Pupil max” (100) and “Model sensitivity” (0.9970) values, which were kept consistent across participants.

3. The subdirectory "gazecalibration/000/" contains the video-based eye-tracking calibration data recorded with Pupil Labs software. Files include:

3.1. eye0.mp4, eye0_timestamps.npy: Eye video files of the participant's right eye during the calibration procedure involving different eye movements. The video was later used with offline pupil detection to calibrate the 3D model of the eye for analysing the pupil images.

3.2. info.csv: Participant session-specific metadata file containing recording information for each session. “Recording Name”(e.g. SP001), “Start Date” (format: DD.MM.YYYY), “Start Time” (format: HH:MM:SS), “Start Time (System)” (format: UNIX epoch time in sec), “Start Time (Synced)”(format: sec relative to the start of the Pupil Capture software's internal time base), “Recording UUID”, “Duration Time” (format: HH:MM:SS), “World Camera Frames” (“0”) and World Camera Resolution (“1280x720” in pixels), “Capture Software Version” (1.15.71), “Data Format Version” (1.15.71), and finally “System Info” giving information on the Computer that the software was run on.

3.3. gaze.pldata, gaze_timestamps.npy, pupil.pldata, pupil_timestamps.npy: Supplementary files containing time-series data on pupil detection gaze and ensuring structural completeness and compatibility of the directory with Pupil Labs software (recorded with version 1.15.71).

3.4. world.intrinsics, world_timestamps.npy: These files act as a placeholder to ensure structural completeness and compatibility of the directory with Pupil Labs software (recorded with version 1.15.71) but do not contain any meaningful data. 3.5. user_info.csv: This placeholder csv file ensures structural completeness and compatibility of the directory with Pupil Labs software (recorded with version 1.15.71).