29
Distributed Mobile Graphics for Windows Phone JIŘÍ DANIHELKA

Distributed Mobile Graphics for Windows 10 Mobile

Embed Size (px)

Citation preview

Page 1: Distributed Mobile Graphics for Windows 10 Mobile

Distributed Mobile Graphics for Windows Phone

JIŘÍ DANIHELKA

Page 2: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 2 Jiří Danihelka - Distributed Mobile Graphics

Motivation

Rendering of Facial Models

Collaborative Device-to-Device Video Streaming

Virtual Cities on Mobile Devices

Conclusion

OVERVIEW

Page 3: Distributed Mobile Graphics for Windows 10 Mobile

MOTIVATION

Page 4: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 4 Jiří Danihelka - Distributed Mobile Graphics

Research Interest

ComputerGraphics

Communication

Distributed mobile graphics

Page 5: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 5 Jiří Danihelka - Distributed Mobile Graphics

limited resources available (memory, CPU, battery)

unstable and paid wireless network connection

sandbox environment for applications – limited access to hardware

different usage scenarios – in use while on the move

additional sensors (accelerometer, GPS, camera)

Mobile Graphics

Page 6: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 6 Jiří Danihelka - Distributed Mobile Graphics

GOALS:1. Bring previously not possible graphical scenarios

to mobile devices despite their limited resources2. Focus on advanced wireless communication

architectures for mobile graphicsMETHODOLOGY: Use validation using mathematical proof,

implementations and measurements

Goals

Page 7: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 7 Jiří Danihelka - Distributed Mobile Graphics

Mobile computing is a mass market Still limited resources (bandwidth, memory, …) Challenge for collaborative applications

Mobile Graphics - Characteristics

New technologies emerge

2005: Embedded systems 2010: Smartphones 2015: Wearables

Published 3 technical papers about wearables since submission of my thesis

Page 8: Distributed Mobile Graphics for Windows 10 Mobile

RENDERING OF FACIAL MODELS

Page 9: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 9 Jiří Danihelka - Distributed Mobile Graphics

Key-frame interpolation animation Key-frame models in face animation = visemes

Page 10: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 10 Jiří Danihelka - Distributed Mobile Graphics

Traditional reduction methods

Page 11: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 11 Jiří Danihelka - Distributed Mobile Graphics

Viseme reduction

How does it work– find similar visemes

(e.g. “f” and “th”)– merge them together

Problems– How to find similar visemes?– How to merge them optimally?

Page 12: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 12 Jiří Danihelka - Distributed Mobile Graphics

How to find similar visemes? Define a metric for visemes

– distance between two models A and B

vA,1

vA,2

vA,3

vB,1

vB,2

vB,3

Page 13: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 (13) Jiří Danihelka - Distributed Mobile Graphics

Semantic Reduction of Face Models

Proof that outputs of both our algoritms – viseme merging algorithm– viseme reduction (clustering) algorithm

are optimal for selected metrics

Reduced resources– saved 38% memory– 2.25 times faster startup

Page 14: Distributed Mobile Graphics for Windows 10 Mobile

COLLABORATIVE DEVICE-TO-DEVICE VIDEO STREAMING

Page 15: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 15 Jiří Danihelka - Distributed Mobile Graphics

Saving cellular data

Page 16: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 16 Jiří Danihelka - Distributed Mobile Graphics

collaborative downloading can save– up to 40% data for video streaming, 25% on average

(real-life experiment at ETH Zurich)– even more for less time-sensitive data (updates, RSS feeds)

Video dissemination strategy

playing buffered partially buffered not buffered

Page 17: Distributed Mobile Graphics for Windows 10 Mobile

VIRTUAL CITIES ON MOBILE DEVICES

Page 18: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 18 Jiří Danihelka - Distributed Mobile Graphics

Generating on demand

Page 19: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 19 Jiří Danihelka - Distributed Mobile Graphics

Generating on demand

Page 20: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 20 Jiří Danihelka - Distributed Mobile Graphics

Generating on demand

Page 21: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 21 Jiří Danihelka - Distributed Mobile Graphics

Traditionally– generators have to respect previous results– geometry generated so far– state of the other generators

Our approach– generators share only the initial seed– they do not have to synchronize their states– that is why we call the method Stateless generation– delivers consistent results regardless of the starting position– generated cities have no size limits – pseudo-infinite

Stateless generation

Page 22: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 22 Jiří Danihelka - Distributed Mobile Graphics

Delaunay triangulation

Our approach

Page 23: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 23 Jiří Danihelka - Distributed Mobile Graphics

Lot generation

Stateless generation

Page 24: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 24 Jiří Danihelka - Distributed Mobile Graphics

Variations of street layout

Stateless generation

Page 25: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 25 Jiří Danihelka - Distributed Mobile Graphics

Final result

Page 26: Distributed Mobile Graphics for Windows 10 Mobile

CONCLUSIONS

Page 27: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 27 Jiří Danihelka - Distributed Mobile Graphics

Bring previously not possible graphical scenarios to mobile devices despite their limited resources– Viseme-reduction reduces

memory requirements

– Stateless-generation generates only visible buildings

– Stateless-generation creates buildings locally on clients and greatly reduces data download

Conclusions

Page 28: Distributed Mobile Graphics for Windows 10 Mobile

11.12.2015 11:00 28 Jiří Danihelka - Distributed Mobile Graphics

Focus on advanced wireless communication architectures for mobile graphics– Discuss and compare architecture

for 3D head applications

– Proposed distributed collaborative video streaming

– Architecture used in stateless-generation method allows synchronization of generated cities (even when not always connected)

Conclusions

Page 29: Distributed Mobile Graphics for Windows 10 Mobile

Jiří Danihelka - Distributed Mobile Graphics

Thanks for your attention!Jiří Danihelka

[email protected]

29