Skip to main content

Python library and CLI for estimation of gaze duration error.

Project description

Gaze Duration Error

Gazerr is an application for estimating the expected error in a gaze duration measurement derived from repeated application of a point of gaze model. It is particularly applicable to machine learning models that work with device cameras to predict a stream of gaze fixation points from facial images.

The method requires an input dataset of validation points from the point of gaze predictive model. This data is used to generate the probability distribution of true gaze durations given a measured gaze duration.

Installation

Install from source code or from PyPi

Usage

The application can be used from the command line by passing in a path to the calibration file and the parameters for the duration measurement that will be bounded. Note: that the penultimate two parameters should be comma separated sets of integers that depict x,y coordiantes in pixels. The measurement length and session lenth should be expressed in milliseconds.

The final parameter is a path to a directory in which to store the raw results

gazerr <CALIBRATION> <MEASUREMENT> <SESSION> <TARGET TOP LEFT> <TARGET BOTTOM RIGHT> <RESULTS>

To use the application without installing it you can employ the runner script. Example below, using the supplied calibration data:

mkdir results/MREC_MAE_50
python ./gazerr-runner.py data/validation_50_MAE.csv 400 1000 40,40 340,290 results/MREC_MAE_50

Alternatively, you may inspect the code and use the library functions directly inside your own application.

Experiments

All experiments for the research paper can be executed via a series of scripts.

Create synthetic calibration data by running

python scripts/generate_datasets.py

Then execute the gazerr exeriments with the following two commands:

scripts/RUN_EXPERIMENTS.sh
scripts/RUN_BIAS_EXPERIMENTS.sh

Finally, analyse the results and generate the plots with

scripts/ANALYSE_RESULTS.sh

Documentation

Additional documentation to be made available on Read the Docs

If you use gazerr in your research please cite the following article

@article{hawkins2022,
   author = {John Hawkins},
   year = {2022},
   title = {Estimating Gaze Duration Error from Eye Tracking Data},
   journal = {TBC}
}

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

gazerr-0.1.1.tar.gz (6.4 kB view details)

Uploaded Source

File details

Details for the file gazerr-0.1.1.tar.gz.

File metadata

  • Download URL: gazerr-0.1.1.tar.gz
  • Upload date:
  • Size: 6.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.2.0 pkginfo/1.5.0.1 requests/2.25.0 setuptools/51.1.0 requests-toolbelt/0.8.0 tqdm/4.61.2 CPython/3.6.10

File hashes

Hashes for gazerr-0.1.1.tar.gz
Algorithm Hash digest
SHA256 a42f8c5c0b74bbdc08060b6141f6122f721a7a88d27f511f73206f9ac94ce9ac
MD5 b6d5d7395b414750182307ac93e86e02
BLAKE2b-256 2e199881fed0061cd69fcaa26d9d7b454136f9106355d75307cded984b41c1b6

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page