docs/index.md


1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

---
title: What is ArGaze?
---

# Enable modular gaze processing pipeline

**Useful links**: [Installation](installation.md) | [Source Repository](https://git.recherche.enac.fr/projects/argaze/repository) | [Issue Tracker](https://git.recherche.enac.fr/projects/argaze/issues) | [Contact](mailto:achil-contact@recherche.enac.fr)

**ArGaze** python toolkit provides a set of classes to build custom-made gaze processing pipelines that works with any kind of eye tracker devices.

![AGaze pipeline](img/argaze_pipeline.png)

## Gaze analysis pipeline

Whether in real time or in post-processing, **ArGaze** provides extensible modules library allowing to select application specific algorithms at each pipeline step:  

* **Fixation/Saccade identification**: dispersion threshold identification, velocity threshold identification, ...  
* **Area Of Interest (AOI) matching**: focus point inside, deviation circle coverage, ...
* **Scan path analysis**: transition matrix, entropy, exploit/explore ratio, ...

Once incoming data formatted as required, all those gaze analysis features can be used with any screen-based eye tracker devices.

[Learn how to build gaze analysis pipelines for various use cases by reading user guide dedicated section](./user_guide/gaze_analysis_pipeline/introduction.md).

## Augmented reality pipeline

Things goes harder when gaze data comes from head-mounted eye tracker devices. That's why **ArGaze** enable 3D modeled **Augmented Reality (AR)** environment description including **Areas Of Interest (AOI)** mapped on <a href="https://docs.opencv.org/4.x/d5/dae/tutorial_aruco_detection.html" target="_blank">OpenCV ArUco markers</a>.

![AR environment axis](img/ar_environment_axis.png)

This AR pipeline can be combined with any wearable eye tracking device python library like Tobii or Pupill glasses.

!!! note
   
      *AR pipeline is greatly inspired by [Andrew T. Duchowski, Vsevolod Peysakhovich and Krzysztof Krejtz article](https://git.recherche.enac.fr/attachments/download/1990/Using_Pose_Estimation_to_Map_Gaze_to_Detected_Fidu.pdf) about using pose estimation to map gaze to detected fiducial markers.*