Presentation for the paper "Evaluation of Multi-Platform Mobile AR Frameworks for Roman Mosaic Augmentation" at the 16th EUROGRAPHICS Workshop on Graphics and Cultural Heritage (EG GCH), Vienna, Austria, 2018.

1. Evaluation of Multi-Platform
Mobile AR Frameworks for
Roman
Mosaic Augmentation
EUROGRAPHICS Workshop on Graphics and Cultural Heritage (2018)
Vienna, Austria, November 12-15, 2018
Jorge C. S. Cardoso, André Belo
CISUC/DEI, Universidade de Coimbra

2. Contents
1. “Mosaico - Conímbriga e Sicó” project
a. Digital Tools for Exploring Roman Mosaic
2. Prototype AR Application
3. AR Framework Testing
4. Conclusions

3. “Mosaico -
Conímbriga e
Sicó” Project

4. “Mosaico - Conímbriga e Sicó” Project
● Based on the Roman Mosaic
Heritage present in the
geographical axis constituted by
○ the Ruins of the Roman city of
Conímbriga,
○ the Roman Villa of Rabaçal, and
○ the Monumental Complex of
Santiago da Guarda.

5. “Mosaico - Conímbriga e Sicó” Project
● Promotes cultural and creative
activities within the museums,
interpretative centers and
archaeological sites
○ Integrated into the CREATOUR
national project as a pilot initiative
● Alternative experiences of
sharing knowledge about the
Roman Mosaic Heritage
● Mosaic as a modern expression
of creativity brought into the
present and reinterpreted

6. Digital Tools for
Exploring
Roman Mosaic
In the context of the “Mosaico -
Conímbriga e Sicó” project
Various planned digital tools to
support different activities for
exploring roman mosaic

7. Digital Tools for Exploring Roman Mosaic
● Mosaic Editor
○ Support for mosaic workshops for
non-professionals
● Programming Environment
(Adapted Snap! Environment)
○ Learning computer programming by
creating mosaic patterns
● Interactive Sandbox
○ Interactive experiences for kids
○ Simulating uncovering mosaics

8. Augmented
Reality Mobile
Application
An AR application that provides
in-place, contextual information
about the roman mosaics

9. Augmented Reality Mobile Application for Mosaics
● Purpose
○ Display technical information about the
mosaics, for example, when they were
uncovered, what was the latest
conservation or restoration work, etc.
○ Display image overlays of the
conservation or restoration works on
mosaics over time.
○ Provide a platform for the visualization of
virtual restoration of the existing mosaics.
○ Highlight mosaics with graphical
information regarding various motifs
■ geometric patterns, animals, plants,
compositions, mythological figures, etc.

10. Types of Augmented Reality
Location-based Structured markers
“Instant tracking” Natural image features

11. Augmented Reality Mobile Application
● Requirement
○ Multi-platform application (run on
Android, iOS, etc.)
■ Single code base
■ Lower development effort
● What AR development
frameworks are available for
multi-platform mobile
development?
● Which AR development
frameworks are most suitable
for detecting real mosaics?

12. AR Framework
Testing

13. AR Development Frameworks
● We studied multi-platform AR development frameworks and their
features
○ With natural image feature detection
● Narrowed down to 3 frameworks: CraftAR, PixLive, Wikitude

14. AR Framework Evaluation
● Real-world evaluation with roman mosaics at Conímbriga
● Targets with different characteristics were captured

15. AR Test Application
● A simple test application was developed using each of the 3 AR
frameworks
● Test application overlaid graphical shapes over the targets
○ Simple shapes, easy to verify alignment

16. AR Test Application
● We screen-captured in video the execution of the application over
each of the mosaic targets
● Three camera movements:
○ Camera face down, turn up towards the target, then turn left, then right
○ Horizontal pan left/right
○ “Zoom in/out”

17. AR Test Application
● We screen-captured in video the execution of the application over
each of the mosaic targets
● We analysed the various videos and extracted 3 metrics
○ Recognition delay
○ Minimum required target area
○ Visual alignment and stability

18. AR Framework Evaluation Results - Overall Recognition
● Not all targets were
recognized
○ This was expected
○ Targets were captured from a
distance
○ Not much effort in capturing
targets
● Wikitude performed very
poorly
○ Unexpected
○ Requires further study as to
why

19. AR Framework Evaluation Results - Recognition Delay
● CraftAR is faster
than PixLive
○ Almost 0.5
seconds faster
● PixLive requires less
visible target area

20. AR Framework Evaluation Results - Visual alignment and
stability
1. Subjectively rated by the two authors
a. -1: bad alignment / stability
b. 0: ok alignment / stability
c. 1: good alignment / stability

21. Conclusion

22. Conclusion
● Evaluation in a real scenario setting
○ Compared three frameworks: CraftAR, PixLive, Wikitude
○ Wikitude failed, but more testing is required to dismiss it
● Study allowed us to understand strong and weak points of these
AR frameworks
○ AR frameworks’ performance varies greatly depending on the type of
image they are recognizing
○ AR frameworks have different performance compromises
■ No single one is best at every performance attribute
● Virtual Heritage application developers should test different
frameworks before commiting to one