The human side of multimedia systems
1 like258 views
The document discusses research topics in multimedia systems, focusing on areas like crowdsourcing, medical video analysis, serious games, and parallel processing. It highlights the potential of gamification in user studies and the challenges in analyzing medical videos, with an emphasis on future directions such as real-time analysis and social event segmentation. The author presents current work and outlines plans for continued research and collaboration in the field.
The human side of multimedia systems
- 1. THE HUMAN SIDE OF MULTIMEDIA SYSTEMS MICHAEL RIEGLER http://goo.gl/uU5J6h
- 2. Outline Research topics What we did so far Future work
- 3. Overview research topics Crowdsourcing Medical Video Analysis Serious Games Parallel Processing Social Computing
- 4. Parallel processing Real time processing of huge multimedia workloads Switch from batch to near realtime Videos and large scale image datasets
- 5. Crowdsourcing Use humans for tasks that computers can not do Can help in many research areas An easy way to perform user studies But can be complicated
- 6. Serious games and gami鍖cation Rising star in the multimedia community Try to use game mechanics for collecting data from users Gamify boring tasks You can be creative! Very interesting for MHealth
- 7. Social event detection Event segmentation Event synchronisation Event classi鍖cation Rising 鍖eld because of the huge amount of collected data from the users
- 8. Medical video analysis Can help to make tasks for doctors much easier Very challenging to get data Analysis of the videos is a challenging task too Quit unexplored area (apart from colonoscopy)
- 9. What we did so far Intentional framing theory Perception of deception Annotation and analysis of medical videos Annotation of game videos Social event synchronisation Serious game for eye tracking Gami鍖cation of cancer screening Crowdsourcing Parallel processing (just started)
- 10. Intentional framing theory Analyse multimedia content based on the creators intent New vista in multimedia research Watch my ACM HRHR talk
- 11. Perception of deception How manipulated images are used online Detect them automatically
- 12. Annotation and analysis of medical videos Annotation tools Hand drawn annotation Semi supervised annotation Endoscopic surgery events Analysis: object tracking, etc.
- 14. SEMI SUPERVISED ANNOTATION TOOL
- 15. EVENT MODEL
- 16. Social event synchronisation Synchronisation of multimedia collections Based on global image features MediaEval participation Best results
- 17. Annotation of video games Like a tool for sportscaster VideoJot Combines zooming (Axel Carlier), drawn annotation and like lines (Raynor Vliegendhart)
- 18. Serious game for eye tracking A serious game as an substitute for eye tracking First studies showed that it works
- 19. Gami鍖cation of cancer screening Proposal for preliminary research Increase the awareness for cervical and breast cancer screening Together with the software engineering group
- 20. Crowdsourcing Gami鍖cation of crowdsourcing Improving the quality Save costs Intrinsic vs. extrinsic motivation
- 21. Parallel processing Explore state of the art Try di鍖erent frameworks Collect data for a large scale experiment (related to all other research topics)
- 22. And we did some art
- 23. WHAT COMES NEXT
- 24. Geo location estimation for large scale data processing Real time analysis of videos GeoMM community gets bigger and bigger Plan is to establish it as an own research 鍖eld Can be used for many systems: Real time recommendation Location based video guide etc.
- 25. Social Event Segmentation Extend our work from MediaEval Event classi鍖cation based on global image features
- 26. Gami鍖cation of cancer screening Wait for the proposal Develop a gami鍖ed app for it
- 27. Perception of Deception Detecting Manipulated images and how they are used online Dataset creation almost 鍖nished Next step: Analysis Create a own MediaEval task
- 28. Serious Game for Eyetracking Very good feedback at ACM MM Next step: compare with real eye tracking Possibility for a collaboration with University of Florida
- 29. Permanent Crowdsourcing Multimedia System Possible collaboration with Vincent Carvillat and Axel Carlier To test multimedia systems (like annotation tools etc.) Never ending crowdsourcing for semi supervised learning
- 30. Continue with intentional framing Speci鍖cation of framing categories User studies Use of computer vision
- 31. THANK YOU