QValiData Experiment Data

Data set for validation experiments performed using the newly-developed QValiData video validation and simulation software. This archive contains:

Data sets for all five video validation trials

R scripts for generating figures used in the manuscript

Final results, organized in CSV form

Directory Structure:

experiment-data

test*

Contains QValiData configuration file, activity logger data, and video recordings for all tests involved in the experiment

post-process

data: Contains data files describing activity detector performance. Each file name is given in the following format: [bit|val]_halfsecond_[nowakeup|wakeup|]_[trial_number]_[activeevents|videoevents]

[bit|val]

bit: Data set uses "Bit Tag" activity detection parameters, developed prior to the implementation of the new methodology.

val: Data set uses parameters found experimentally using simulation-based validation.

halfsecond: The "ground-truth" data set used (list of events classified as "active" by video) was trimmed to only include events lasting half a second or longer (15 or more frames of video)

[nowakeup|wakeup]

nowakeup: ADXL362 low-power activity wakeup mode was not used for this trial.

wakeup: wakeup mode was used for this trial.

[trial_number]: A number in the range 1-5 indicating which data set the results came from

[activeevents|videoevents]

activeevents: The file contains a list of all events detected by the simulated activity detector. Each row contains a number indicating the length of each event (in seconds) and a TRUE/FALSE value that is TRUE only if the event contains at least one sample that was also marked as "active" by video analysis.

videoevents: The file contains a list of all events detected by video analysis. Each row contains a number indicating the length of each event (in seconds) and another number indicating the percentage of each trial that was also marked as "active" by the simulated activity detector.

correctevents.R: Plots histograms of events detected via video analysis that were also detected by simulated activity detection using each set of activity detection parameters.

falsepositives.R: Plots histograms of events detected via simulated activity detection, but were not detected by video analysis, which reveals the occurrence of false-positive classifications.

post-process.Rproj: R project file for both R scripts.

QValiData-Parameters.csv: Contains all activity detection parameters used for the five trials.

QValiData-Results.csv: Contains simulation performance metrics for all five trials.

This material is based upon work supported by the National Science Foundation under Grant No. 1644717.