Smart Glasses & Applications Presenter: Xu Qiu Advisor: C.-C. (Jay) Kuo 1 1
Outline Introduction Comparison of Smart Glasses Apps for Developing Platform Development and Implementation of Computer Vision Algorithms 2 1
Outline Introduction - Project Introduction - GE90 Jet Engine Repair Comparison of Smart Glasses Apps for Developing Platform Development and Implementation of Computer Vision Algorithms 3 1
From May 2014 http://mcl.usc.edu/news/2014/04/20/media-communications-lab-beginsresearch-projects-with-google-glass/ 4
GE90 Jet Engine Repair https://www.youtube.com/watch?v=ndkqo0pzmqm 5
GE90 Jet Engine Repair Recognize part 6
GE90 Jet Engine Repair Recognize part 7
GE90 Jet Engine Repair Engine status 8
GE90 Jet Engine Repair Recognize someone 9
GE90 Jet Engine Repair Pair with Boroscope 10
GE90 Jet Engine Repair Get directions to 11
GE90 Jet Engine Repair Send/get messages 12
GE90 Jet Engine Repair Task list 13
Outline Introduction Comparison of Smart Glasses - VS Vuzix M100 - Hardware Apps for Developing Platform Development and Implementation of Computer Vision Algorithms 14 1
Comparison Smart Glasses VS Vuzix M100 Vuzix M100 Smart Glasses 15
Vuzix Smart Glasses Display An image visually equivalent to a 4 in. smartphone screen held at a typical 14 in. distance. 16
Hardware Display: Prism projector, 640 360 pixels (equivalent of a 25 in/64 cm screen from 8 ft/2.4 m away) Sound: Bone conduction transducer Input: Voice command through microphone, accelerometer, gyroscope, magnetometer, light sensor, proximity sensor Controller input: Touchpad, MyGlass phone app Camera: Photos 5 MP, videos 720p Connectivity: Wi-Fi, Bluetooth, micro USB Weight: 50g Backward compatibility: - Any Bluetooth-capable phone; - MyGlass app requires Android 4.0.3 + or any ios 7.0 + 17
Hardware 18
Hardware Touchpad: 19
Outline Introduction Comparison of Smart Glasses Apps for - Social network - GDK games - LBS(GPS apps) - Detection/Recognition Apps - Audio apps - Other apps Developing Platform Development and Implementation of Computer Vision Algorithms 20 1
Apps: Social network http://glass-apps.org/google-glass-application-list 21
Apps: GDK games: Missile Beyond http://glass-apps.org/google-glass-application-list 22
Apps: GDK games: GlassCopter http://glass-apps.org/google-glass-application-list 23
Apps: LBS(GPS apps): AR Glass for Wikipedia http://glass-apps.org/google-glass-application-list 24
Apps: SpeedHud http://glass-apps.org/google-glass-application-list 25
Apps: Detection/Recognition Apps http://glass-apps.org/google-glass-application-list 26
Apps: Audio apps: Name This Song: Ok glass, name this song. GlassMusic: search by any song name, artist or album. Other apps: Speech Helper for Glass: Upload your speech draft and slides. Check a battery: check a battery for. http://glass-apps.org/google-glass-application-list 27
Outline Introduction Comparison of Smart Glasses Apps for Developing Platform - GDK - Mirror API Development and Implementation of Computer Vision Algorithms 28 1
Developing platform GDK: Real-time user interaction, offline functionality, access to hardware. 29
Developing platform Mirror API: platform independence, common infrastructure, built-in functionality 30
Outline Introduction Comparison of Smart Glasses Apps for Developing Platform Development and Implementation of Computer Vision Algorithms - Relative Viewing Angle Estimation - Future work 31 1
Development and Implementation of Computer Vision Algorithms: Visual Saliency Determination Relative Viewing Angle Estimation Object Detection 32
Relative Viewing Angle Estimation Moving camera instead of smart glasses Real-time algorithms (CVPR 2013)Real-time Model-based Rigid Object Pose Estimation and Tracking Combining Dense and Sparse Visual Cues, by Karl Pauwels, Leonardo Rubio, Javier D íaz, Eduardo Ros of University of Granada, Spain. 33
Relative Viewing Angle Estimation Real-time Model-based Rigid Object Pose Estimation and Tracking Combining Dense and Sparse Visual Cues Estimating and tracking the six-degrees-of-freedom (6DOF) pose of rigid objects of arbitrary shapes in real-time. - highly accurate and robust to noise and occlusions Problems: -Stereo Camera: -Graphics Processing Units (GPUs) (CVPR 2013)Real-time Model-based Rigid Object Pose Estimation and Tracking Combining Dense and Sparse Visual Cues, by Karl Pauwels, Leonardo Rubio, Javier D íaz, Eduardo Ros of University of Granada, Spain. 34
Relative Viewing Angle Estimation Moving camera instead of smart glasses Real-time algorithms No stereo-camera Multi-target tracking with Single Moving Camera - multi-target tracking - unknown camera motion - did not rely on the use of stereo-camera. http://www.eecs.umich.edu/vision/mttproject.html Object detection: Felzenszwalb, P.F., Girshick, R.B., McAllester, D., Ramanan, D.: Object detection with discriminatively trained part based models. In: PAMI. (2009) 35
Relative Viewing Angle Estimation Multi-target tracking with Single Moving Camera 36
Future work Relative Viewing Angle Estimation - Limitation: Single camera Storage, Computation capacity - Challenges: Moving camera Unknown background - Possible Solutions: Cloud Simply the algorithms Use motion to get 3D information 37
Thank you Comments or Questions? 38