ProsumerFX: Mobile Design of Image Stylization Components

ProsumerFX:
Mobile Design of Image Stylization Components
Tobias Dürschmid, Maximilian Söchting, Amir Semmo, Mattias Trapp, Jürgen Döllner
Hasso Plattner Institute, University of Potsdam, Germany
• Large amount of casual creativity apps that allow the application of
preconfigured visual styles to images
• Filtered images more likely to be shared on social media [Bakhshi et al. 2015]
MOTIVATION
Instagram Prisma
https://commons.wikimedia.org/wiki/File:Instagram_collage_with_15_different_filters.jpg CC BY 3.0
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 2
MOTIVATION
?
Existing approaches (Consumption)
Presented approach (Prosumption)
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 3
• Goal: Enable prosumption of visual styles for non-expert users
• Have users compose own visual styles with easy-to-use building blocks
• Have users share own visual styles through a web-based platform
MOTIVATION
Prosumption of visual styles
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 4
• Enable rapid prototyping and A/B testing of visual styles
• Modification of the presentation of the created visual styles
MOTIVATION
Prototype
Review
Design &
Iterate
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 5
DEMO
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 6
https://www.youtube.com/watch?v=yv8N1tIce84
• Imageplay.io: Allows combination of low-level GPU-based algorithms
• Snapchat Geofilters: Allows building of basic transparency-based filters
with texts and images
RELATED WORK
Imageplay Snapchat Geofilters
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 7
• Interactive Modification
• User-friendly UI for various modifications
• Modifications have to be propagated to the GPU immediately
• Serialization of visual styles
• Persistence of user-defined visual styles
• Device Heterogeneity
• Many different mobile and desktop devices
• Reusable Building Blocks (e.g., edge detection, color quantization)
• Modularization of image stylization components
CHALLENGES
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 8
Modifications of parameters,
presets, and effects:
• Swap
• Remove
• Rename
• Exchange icon
CONCEPT – UI
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 9
CONCEPT – INTERACTION
Level of Control
[Semmo et al. 2016]
Consumption
Prosumption
low medium high
Parameter
Modification
Preset
Modification
Effect
Recombination
Preset
Selection
Parameter
Adjustment
Parameter
Painting
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 10
CONCEPT – STYLE SERIALIZATION
• Store and transfer visual styles
in a serialized form
• An effect consists of
• Rendering passes
• Shader programs
• Textures
• Parameters
• Presets
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 11
CONCEPT – PLATFORM INDEPENDENCE
Effect
Definition
Effect
Implementation
OpenGL ES 2.0
Effect
Implementation
OpenGL 4.5
(Desktop)
Parameters
Presets
Icons
Shaders
Rendering Passes
Define the platform-
independent part
(user modifiable)
Define the platform-
specific part
(not user modifiable)
Effect
Definition
Effect
Implementation
OpenGL ES 2.0
(Android, iOS)
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 12
CONCEPT – REUSABLE BUILDING BLOCKS
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 13
• Exploration of all publicly available visual styles in a web store-like
interface
• Publishing of newly created visual styles through a REST API
CONCEPT – SERVER
Alternative views of the web interface
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 14
IMPLEMENTATION
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 15
RESULTS
Building blocks:
• Neural Style Transfer
[Johnson et al. 2016]
• Toon Filtering
[Winnemöller et al.
2006]
• Watercolor
[Bousseau 2006]
• Color grading
[Selan 2004]
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 16
RESULTS
Building blocks:
• Neural Style Transfer
[Johnson et al. 2016]
• Toon Filtering
[Winnemöller et al.
2006]
• Watercolor
[Bousseau 2006]
• Color grading
[Selan 2004]
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 17
RESULTS
Building blocks:
• Neural Style Transfer
[Johnson et al. 2016]
• Toon Filtering
[Winnemöller et al.
2006]
• Watercolor
[Bousseau 2006]
• Color grading
[Selan 2004]
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 18
RESULTS
Building blocks:
• Neural Style Transfer
[Johnson et al. 2016]
• Toon Filtering
[Winnemöller et al.
2006]
• Watercolor
[Bousseau 2006]
• Color grading
[Selan 2004]
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 19
RESULTS
Building blocks:
• Neural Style Transfer
[Johnson et al. 2016]
• Toon Filtering
[Winnemöller et al.
2006]
• Watercolor
[Bousseau 2006]
• Color grading
[Selan 2004]
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 20
RESULTS
Building blocks:
• Neural Style Transfer
[Johnson et al. 2016]
• Toon Filtering
[Winnemöller et al.
2006]
• Watercolor
[Bousseau 2006]
• Color grading
[Selan 2004]
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 21
RESULTS
Building blocks:
• Neural Style Transfer
[Johnson et al. 2016]
• Toon Filtering
[Winnemöller et al.
2006]
• Watercolor
[Bousseau 2006]
• Color grading
[Selan 2004]
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 22
RESULTS
Test Device: Samsung S7 (Mali T880)
Result: No additional processing
overhead when combining effects.
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 23
RESULTS
Test Device: Samsung S7 (Mali T880)
Result: The capacity of effects
depends on the rendering resolution.
WQHD: 9 effects
FHD: 12 effects
HD: 14 effects
Pencil Hatching Effect:
Flow-oriented example-based hatching with
contours, consisting of 14 render-to-texture
passes, five half-float textures, 17 byte textures,
and 18 small tonal art map textures
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 24
• Swapping effects: 1ms + rendering time
• Adding an effect ≈1s + rendering time
• Removing an effect: 2ms + rendering time
Test Device: Samsung Galaxy S7 (CPU octa-core 4x2.3 GHz, 4x1.6 GHz)
Result: Style composition is interactive.
RESULTS – PERFORMANCE
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 26
• Creation of visual styles is enabled in real-time
• Support for sharing visual styles via XML-based document format
• Device heterogeneity achieved through platform-independent
persistence
• Proof of concept: effect database contains 190 effects (e.g., neural style
transfer, water color, oil paint, cartoon, pencil hatching)
DISCUSSION
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 27
CONTRIBUTIONS
• Concept for mobile prosumption of image stylization components
• Document format for platform-independent persistence of visual styles
• Web-based platform for sharing visual styles
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 28
FUTURE WORK
• Enable low level effect modifications
• Exchange processing steps of effect
• Exchange of textures
• Collapsing similar parameters
• Hinting rendering performance
• Community platform
• Public release
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 29
CONCLUSION
• Enable novice users and professionals to create and manipulate their
individual visual styles using one ecosystem
• Eases rapid prototyping & A/B testing of visual styles for mobile platforms
• The presented concept goes beyond user-generated content,
because users generate tools (visual styles) that enable other users to
generate content (styled images)
27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 30
THANK YOU FOR YOUR ATTENTION
Tobias Dürschmid: tobias.duerschmid@student.hpi.de
Maximilian Söchting: maximilian.soechting@student.hpi.de
Amir Semmo: amir.semmmo@hpi.de
Matthias Trapp: matthias.trapp@hpi.de
Jürgen Döllner: juergen.doellner@hpi.de
Project 01IS15041B
1 of 30

More Related Content

Recently uploaded(20)

Featured(20)

ChatGPT webinar slidesChatGPT webinar slides
ChatGPT webinar slides
Alireza Esmikhani30.3K views
More than Just Lines on a Map: Best Practices for U.S Bike RoutesMore than Just Lines on a Map: Best Practices for U.S Bike Routes
More than Just Lines on a Map: Best Practices for U.S Bike Routes
Project for Public Spaces & National Center for Biking and Walking6.9K views
Barbie - Brand Strategy PresentationBarbie - Brand Strategy Presentation
Barbie - Brand Strategy Presentation
Erica Santiago25.1K views
9 Tips for a Work-free Vacation9 Tips for a Work-free Vacation
9 Tips for a Work-free Vacation
Weekdone.com7.2K views
I Rock Therefore I Am. 20 Legendary Quotes from PrinceI Rock Therefore I Am. 20 Legendary Quotes from Prince
I Rock Therefore I Am. 20 Legendary Quotes from Prince
Empowered Presentations142.8K views
How to Map Your FutureHow to Map Your Future
How to Map Your Future
SlideShop.com275.1K views
Read with Pride | LGBTQ+ ReadsRead with Pride | LGBTQ+ Reads
Read with Pride | LGBTQ+ Reads
Kayla Martin-Gant1.1K views
The Student's Guide to LinkedInThe Student's Guide to LinkedIn
The Student's Guide to LinkedIn
LinkedIn87.9K views

ProsumerFX: Mobile Design of Image Stylization Components

  • 1. ProsumerFX: Mobile Design of Image Stylization Components Tobias Dürschmid, Maximilian Söchting, Amir Semmo, Mattias Trapp, Jürgen Döllner Hasso Plattner Institute, University of Potsdam, Germany
  • 2. • Large amount of casual creativity apps that allow the application of preconfigured visual styles to images • Filtered images more likely to be shared on social media [Bakhshi et al. 2015] MOTIVATION Instagram Prisma https://commons.wikimedia.org/wiki/File:Instagram_collage_with_15_different_filters.jpg CC BY 3.0 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 2
  • 3. MOTIVATION ? Existing approaches (Consumption) Presented approach (Prosumption) 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 3
  • 4. • Goal: Enable prosumption of visual styles for non-expert users • Have users compose own visual styles with easy-to-use building blocks • Have users share own visual styles through a web-based platform MOTIVATION Prosumption of visual styles 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 4
  • 5. • Enable rapid prototyping and A/B testing of visual styles • Modification of the presentation of the created visual styles MOTIVATION Prototype Review Design & Iterate 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 5
  • 6. DEMO 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 6 https://www.youtube.com/watch?v=yv8N1tIce84
  • 7. • Imageplay.io: Allows combination of low-level GPU-based algorithms • Snapchat Geofilters: Allows building of basic transparency-based filters with texts and images RELATED WORK Imageplay Snapchat Geofilters 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 7
  • 8. • Interactive Modification • User-friendly UI for various modifications • Modifications have to be propagated to the GPU immediately • Serialization of visual styles • Persistence of user-defined visual styles • Device Heterogeneity • Many different mobile and desktop devices • Reusable Building Blocks (e.g., edge detection, color quantization) • Modularization of image stylization components CHALLENGES 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 8
  • 9. Modifications of parameters, presets, and effects: • Swap • Remove • Rename • Exchange icon CONCEPT – UI 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 9
  • 10. CONCEPT – INTERACTION Level of Control [Semmo et al. 2016] Consumption Prosumption low medium high Parameter Modification Preset Modification Effect Recombination Preset Selection Parameter Adjustment Parameter Painting 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 10
  • 11. CONCEPT – STYLE SERIALIZATION • Store and transfer visual styles in a serialized form • An effect consists of • Rendering passes • Shader programs • Textures • Parameters • Presets 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 11
  • 12. CONCEPT – PLATFORM INDEPENDENCE Effect Definition Effect Implementation OpenGL ES 2.0 Effect Implementation OpenGL 4.5 (Desktop) Parameters Presets Icons Shaders Rendering Passes Define the platform- independent part (user modifiable) Define the platform- specific part (not user modifiable) Effect Definition Effect Implementation OpenGL ES 2.0 (Android, iOS) 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 12
  • 13. CONCEPT – REUSABLE BUILDING BLOCKS 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 13
  • 14. • Exploration of all publicly available visual styles in a web store-like interface • Publishing of newly created visual styles through a REST API CONCEPT – SERVER Alternative views of the web interface 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 14
  • 15. IMPLEMENTATION 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 15
  • 16. RESULTS Building blocks: • Neural Style Transfer [Johnson et al. 2016] • Toon Filtering [Winnemöller et al. 2006] • Watercolor [Bousseau 2006] • Color grading [Selan 2004] 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 16
  • 17. RESULTS Building blocks: • Neural Style Transfer [Johnson et al. 2016] • Toon Filtering [Winnemöller et al. 2006] • Watercolor [Bousseau 2006] • Color grading [Selan 2004] 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 17
  • 18. RESULTS Building blocks: • Neural Style Transfer [Johnson et al. 2016] • Toon Filtering [Winnemöller et al. 2006] • Watercolor [Bousseau 2006] • Color grading [Selan 2004] 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 18
  • 19. RESULTS Building blocks: • Neural Style Transfer [Johnson et al. 2016] • Toon Filtering [Winnemöller et al. 2006] • Watercolor [Bousseau 2006] • Color grading [Selan 2004] 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 19
  • 20. RESULTS Building blocks: • Neural Style Transfer [Johnson et al. 2016] • Toon Filtering [Winnemöller et al. 2006] • Watercolor [Bousseau 2006] • Color grading [Selan 2004] 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 20
  • 21. RESULTS Building blocks: • Neural Style Transfer [Johnson et al. 2016] • Toon Filtering [Winnemöller et al. 2006] • Watercolor [Bousseau 2006] • Color grading [Selan 2004] 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 21
  • 22. RESULTS Building blocks: • Neural Style Transfer [Johnson et al. 2016] • Toon Filtering [Winnemöller et al. 2006] • Watercolor [Bousseau 2006] • Color grading [Selan 2004] 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 22
  • 23. RESULTS Test Device: Samsung S7 (Mali T880) Result: No additional processing overhead when combining effects. 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 23
  • 24. RESULTS Test Device: Samsung S7 (Mali T880) Result: The capacity of effects depends on the rendering resolution. WQHD: 9 effects FHD: 12 effects HD: 14 effects Pencil Hatching Effect: Flow-oriented example-based hatching with contours, consisting of 14 render-to-texture passes, five half-float textures, 17 byte textures, and 18 small tonal art map textures 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 24
  • 25. • Swapping effects: 1ms + rendering time • Adding an effect ≈1s + rendering time • Removing an effect: 2ms + rendering time Test Device: Samsung Galaxy S7 (CPU octa-core 4x2.3 GHz, 4x1.6 GHz) Result: Style composition is interactive. RESULTS – PERFORMANCE 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 26
  • 26. • Creation of visual styles is enabled in real-time • Support for sharing visual styles via XML-based document format • Device heterogeneity achieved through platform-independent persistence • Proof of concept: effect database contains 190 effects (e.g., neural style transfer, water color, oil paint, cartoon, pencil hatching) DISCUSSION 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 27
  • 27. CONTRIBUTIONS • Concept for mobile prosumption of image stylization components • Document format for platform-independent persistence of visual styles • Web-based platform for sharing visual styles 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 28
  • 28. FUTURE WORK • Enable low level effect modifications • Exchange processing steps of effect • Exchange of textures • Collapsing similar parameters • Hinting rendering performance • Community platform • Public release 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 29
  • 29. CONCLUSION • Enable novice users and professionals to create and manipulate their individual visual styles using one ecosystem • Eases rapid prototyping & A/B testing of visual styles for mobile platforms • The presented concept goes beyond user-generated content, because users generate tools (visual styles) that enable other users to generate content (styled images) 27.11.2017 ProsumerFX: Mobile Design of Image Stylization Components, T. Dürschmid et al. 30
  • 30. THANK YOU FOR YOUR ATTENTION Tobias Dürschmid: tobias.duerschmid@student.hpi.de Maximilian Söchting: maximilian.soechting@student.hpi.de Amir Semmo: amir.semmmo@hpi.de Matthias Trapp: matthias.trapp@hpi.de Jürgen Döllner: juergen.doellner@hpi.de Project 01IS15041B