Kris M. Kitani

 

Profile

Postdoctoral research fellow in the Robotics Institute Computer Vision Group at Carnegie Mellon University since 2011 with Martial Hebert and Drew Bagnell.


Cooperative research fellow in the Sato Lab,
at the University of Tokyo with Yoichi Sato.


Assistant professor at the University of Electro-Communications from 2008 to 2011 with Hideki Koike.

Visiting scholar at UCSD with Serge Belongie in 2010.


University of Southern California (BS, Electrical Engineering)

University of Tokyo (MS, PhD)

Research - COMPUTER VISION

RECENT NEWS

Activity Forecasting:

Predicting future actions before they happen using physical scene features

(ECCV 2012)

Best Paper Runner-up ECCV 2012

EgoAction Analysis:

Discovering first-person action categories using ego-motion analysis (MIRU 2010, CVPR 2011)

Motion-in-Context:

Discovering action categories using hand motion and interactive objects

(VS 2008, MIRU 2008)

Best Student Paper MIRU 2008

Best Journal Paper IEICE 2010

Learning Activity Grammars:

Discovering the grammar of activities from noisy visual observations

(WMVC 2007, IJPRAI 2008)

Research - HUMAN-COMPUTER INTERACTION

Coupled Gaze & Motion Analysis:

Recognizing first-person actions from gaze and ego-motion (ECV2012)

Image Sonification:

Exploration of visual image features via sonic feedback (AH2011)

Human-centric Image Stitching:

Preserving people in composite panoramic images (AH 2012)

Algorithmic Musical Augmentation:

Stochastic context-free grammars for on-the-fly percussion accompaniment (NIME 2010)

BallCam!

Image processing for spinning cameras (UIST 2012, AH 2013) *

Probabilistic Grammars for Recognition:

Representation and inference over hierarchical structures of human activity (VS-PETS 2005)

BallCam! Demo Video:

  1. 1.FPV hand-detection paper accepted at CVPR 2013.

  2. 2.Media coverage: BallCam project in the NewScientist, Popular Mechanics and Wall Street Journal.

  3. 3.Received Best Paper Award Honorable Mention for Activity Forecasting at ECCV 2012. More information on the Activity Forecasting Project page.

  4. 4.Media coverage: Activity Forecasting at CNET.com, phys.org and petapixel.com

  5. 5.BallCam! paper accepted at UIST 2012 and Augmented Human 2013
    (references updated see below!)

RECENT MEDIA COVERAGE

  1. •“A Football's-Eye View.” Wall Street Journal, Week in Ideas. March 8, 2013.

  2. •“Spinning camera gives a ball's-eye view of the game.” New Scientist, 2013.

  3. •“Pa. Student’s Software Captures The Game From The Ball’s Perspective.” CBS Philly, 2013.

  4. •“BallCam Puts Football Fans on the Field.” DesignNews, 2013.

  5. •“Local Students Create Camera Inside Of A Football.” CBS Pittsburgh (KDKA News at 6), 2013.

  6. •“The CMU Ball Cam: Watch any sport from the perspective of the ball.” PopCity Media, 2013.

  7. •“What the World Looks Like From a Football's Point of View.” Popular Mechanics, 2013.

  8. •“Football Physics: Watching the game from the eye of the football.” Physics Central, 2013.

  9. •“BallCam Utilizes Algorithms And GoPro Camera To Provide New Field Of View.” RedOrbit, 2013.

  10. •“U.S. looks to replace human surveillance with computers.” CNET, 2012.

  11. •“Surveillance tech from Carnegie Mellon can watch and predict.” PHYS.ORG, 2012.

  12. •“Scientists Building Security Cameras That Can “See” Crimes Before They Happen.” PetaPixel, 2012.

Tracking in Crowds:

Using individual appearance and gait frequency to track in crowds

(ICCV 2009)

 

  1. *We note here similar ideas have been implemented or proposed but not included in our camera ready paper. See our extended AH2013 paper for more details.
    1938 RKO-Radio Pictures (placed a video camera in a flying football),
    1992 Sandia Labs [DART project] (rotating image sensor and ground registration)
    2009 S. Hollinger [www.serveball.com] (multi-camera, multi-sensor ball for in-flight imaging)

Ego-Centric Hand Detection:

Robust pixel-wise detection of hands for first-person vision (FPV)

CVPR 2013