10. A brief history of eye-tracking research
10
Long-standing interest in the study of visual attention in
various research disciplines:
Psycholinguistics:
reading research
Psychology:
scene perception
11. A brief history of eye-tracking research
11
A more recent research interest can be observed in:
Human-computer
interaction
Human-human
interaction
12. A brief history of eye-tracking research
12
Navigation &
wayfinding
Kinematics & sports
research
A more recent research interest can be observed in:
16. Developments in eye-tracking technology
16
User-friendly & flexible recording devices are one thing,
but efficient data analysis is a completely different story
17. Developments in eye-tracking technology
17
Recent development:
Efficient data aggregation and analysis tools:
semi-automatic analysis & results presentation
à towards plug and play systems
19. Developments in eye-tracking technology
19
Not ok for mobile eye-trackers
Ø No fixed reference frame (moving
head, moving subject)
Ø Potentially multiple moving objects
in the scene
Ø Highly complex datastream for
mobile eye-tracking
20. Developments in eye-tracking technology
20
• Manual coding
o time-consuming (and thus expensive!)
o requires technical expertise
Current options:
21. Developments in eye-tracking technology
21
• Predefine potentional area of analysis
o Based on infrared (or other) markers
o 2-D plane of zone predefined by markers
o Semi automatic data aggregation & analysis possible
Current options:
23. 23
• But:
o Works only for predefined planes
o Tracking multiple fields or objects with identical or similar features
(object categories)
o Objects of interest need to be tied to a fixed position in the AOA
(<-> handling of objects)
o Labo setup / large natural test environments
Developments in eye-tracking technology
Eye-tracking in the wild?
24. Introducting the InSight Out method
24
• Apply image processing techniques on data collected
by a mobile eye-tracker
• (Semi)-Automatic analysis of (mobile) eye-tracking data
without predefined AOA’s
• User-friendly output generation:
o Time-line
o Statistical data
o Object clouds
o …
25. 25
• Integration of image recognition algorithms
• Benefits:
o Target of analysis is not restricted to a region
o Objects can be moving
o Manual labour limited
Introducting the InSight Out method
26. Basic image recognition techniques
26
Technique #1: Object recognition based on local feature matching
27. Object recognition in eye tracking video
27
User’s
selection
Visual
similarity
score
[ORB: an efficient
alternative to SIFT and
SURF, E. Rublee & G.
Bradski, ICCV 2011 ]
27
31. Varying shape detection
31
• Detection of persons: we trained a new model to detect
upper part of a human body (upper 60% of full body)
o [Parts-based latent SVM cascaded classifier, P. Felzenszwalb, CVPR 2010 ]
• Face detection: 3 face models (frontal, left and right profile)
o [Viola&Jones: Robust Real-time Object Detection, IJCV 2001]
32. Eye-tracker experiments
Experiments used for development and testing
32
• Visiting a library and picking up magazines
• Walking through a public building while paying
attention to signs such as fire exit, staircases
• Walking through the streets while paying
attention to traffic signs
• Visiting a toy shop and picking up products
• Attending a presentation given by a lecturer
33. Eye-tracker experiments
Case study 1: customer journey experiment
33
• “Gain insights in the experience of customers”
• Find relation between user experience and visual behavior
• Experiment was performed in Museum M (Leuven)
• Visiting a specific exhibition: Hieronymus Cock
• 14 participants were involved in this experiments
• 4 systems were used
o Tobii / arrington / contour head mounted camera
• We collected 160GB of data
34. Eye-tracker experiments
Case study 1: customer journey experiment
34
• Questions to be answered:
• Do the visitors notice to walking guides?
• Do the visitor notice the childquiz?
• Do the visitors notice the Ipod / Ipad in the exhibition
• Is there a relation between favorite work and view time?
36. Future developments: near future
36
Attractive visualisations of the detection results
Detection
results
database
• Timeline
• Object statistics
37. 37
• Camera-based localisation and 3D mapping
o Test person location: 2D heat map and location tracks
o 3D gaze location: 3D-heat map
• Possibly combined with detected objects
Future developments: further future
Obj1
Obj2
Obj3
38. 38
• Emotion recognition
o Important aspect of customer journey analysis in UX
o mobile eye-tracker with additional camera which captures the face
o allows to use existing emotion detection algorithms based on the
pose of e.g. mouth corners
o Link with detection of touch points and visualize…
Future developments: quite far future
39. Project planning
39
2 parallel tracks
• PhD research on more theoretical aspects
o Stijn De Beugher
o Sept 2012 - 2016
• Commercial valorisation
o Working towards spin-off startup
o Mission: processing eye-tracking data from experiments conducted
by UX/marketing research bureaus
o Result: nice-looking reports
o Accepting first commercial projects by Q1 2014
guinea pig
discount for first
projects!