Competitor Software

We want to evaluate how Mediapipe compares to other open-source software and computer vision options.

Some opensource options:

xLabs

javascript

webcam

Gaze data is calculated locally and in realtime

Doesn’t require calibration for face detection, but it does for eye gaze.

Outputs a 76-vertex 3D model of the user’s facial features.

Besides tracking where the user is looking on the screen, it can provide details of raw eye movements, which can reveal mental states like tiredness, without calibration.

GazePointer

https://gazerecorder.com/gazepointer/

webcam

c++

calibration needed

application: restricted to control mouse cursor position

Ogama

http://www.ogama.net/

C#

requires infrared webcams

 Its main features include database-driven pre-processing and filtering of gaze and mouse data, the creation of attention maps, areas of interest definition, saliency calculation

openEyes

Can be used with webcams and infrared eye trackers

Requires Matlab

PyGaze

http://www.pygaze.org/category/eyetracking/

Python

PyGaze has been benchmarked on EyeLink, iViewX, and Tobii setups, ensuring compatibility and reliability.

PyGaze supports Windows, Linux, and Mac OSX

Designed to work with eye trackers devices, but there are some ongoing projects with webcam

using PyGame-based code

we will have to tell the software where your pupil is once.

OpenGazer

webcam

Requires Linux

Feature point selection: During startup, the user is expected to select feature points on the face using the mouse.

Calibration required

TurkerGaze

webcam

javascript

Requires Linux

GazeParser / Simple Gaze Tracker

https://link.springer.com/article/10.3758/s13428-012-0286-x

GazeParser consists of two components. GazeParser is a Python module for stimulus presentation, data conversion, and analysis. SimpleGazeTracker is an application for gaze recording written in C++.

Two computers (PC) are necessary to use the SimpleGazeTracker. 

Requires a chinstrap/head restraint to restrict head movement

ITU Gaze Tracker

https://github.com/devinbarry/GazeTracker

webcam

C#

requires calibration

The network API relies on TCP/IP and UDP

Pupil Labs

https://pupil-labs.com/about/

Requires eye tracker devices

They have glasses that don’t require calibration

Open API

EyeRecToo

https://github.com/tcsantini/EyeRecToo

https://www.scitepress.org/Papers/2017/62247/62247.pdf

a second-generation open-source software designed for head-mounted eye trackers.

For data analysis, the software recommends using “Eyetrace,” another tool.

The software has been tested with eye trackers like Dikablis Essential/Pro, Pupil DIY, Eivazi’s microscope add-on, Pupil Eye Tracker (Cam1 and Cam2 models), and inconspicuous modular eye trackers from Eivazi, S et al 2018. It also supports tested webcams like PlayStation Eye, Microsoft LifeCams, Logitech C series, etc.

EyeWriter

https://eyewriter.org/

The EyeWriter was specifically designed to provide individuals with physical disabilities, such as ALS, the ability to communicate and express themselves using eye movements.

The EyeWriter system involves a combination of hardware and software components.

requires calibration

Requires eye tracker devices.

OpenCV

https://towardsdatascience.com/real-time-eye-tracking-using-opencv-and-dlib-b504ca724ac6

It is an open-source computer vision and machine learning software library designed to aid developers in building applications related to computer vision tasks.

OpenCV is cross-platform and works on various operating systems, including Windows, macOS, Linux, Android, and iOS.

OpenCV offers a comprehensive set of functions for image processing, including operations like filtering, edge detection, resizing, color space conversion, and more.

It provides tools for capturing, manipulating, and processing video streams, making it useful for tasks such as real-time video analysis, video enhancement, and tracking.

OpenCV and MediaPipe are two powerful libraries that can be combined to create an eye-tracking system. OpenCV provides image processing and computer vision capabilities, while MediaPipe offers a range of pre-built solutions for various tasks, including facial and landmark detection.

https://www.ijprems.com/uploadedfiles/paper/issue_5_may_2022/309/final/fin_ijprems1657463024.pdf

https://www.youtube.com/watch?v=Y-mCtkv41rk

https://upcommons.upc.edu/handle/2117/384415