コンピュータビジョン特論 - wakayama-u.ac.jp

1 1 Advanced Computer vision 2014 4 A204 3 OpenCV 3 3 10 50% 40% 2 A4 2 HP ~wuhy/ACV8 1.

2 CV( ) CV( ) 3. Nice sunset!CV 2 3 3 3 2 2 Projections or CV 2 3 ModelOutput:Real SceneCamerasImages 3 Scene/object modelingNavigationObject recognitionEvent/action Video AnalysisMultiple frames, temporal informationVideo Understanding3 Computer vision ( )LightingSceneCameraComputerPhysical ModelsScene InterpretationWe need to understand the Geometricand Radiometricrelationsbetween the scene and its Obscura, Gemma Frisius, 15581558 Lens Based Camera Obscura, 156815581568 Still Life, Louis Jaques Mande Daguerre, 1837155818371568 Silicon Image Detector, 19701558183715681970 15581837156819701995 Digital Cameras2010 A Brief History of Images19804 Many new digital cameras now detect faces Canon, Sony, Fuji.

3 :ON :OFF Smile detection?Sony Cyber-shot T70 Digital Still Camera Computer vision CVLightingSceneCameraComputerPhysical ModelsScene InterpretationWe need to understand the Geometricand Radiometricrelationsbetween the scene and its image. 3 2 2 3 3 Computer VisionMake computers understandimages and kind of scene?Where are the cars?How far is the building?..5CV Object categorization skybuildingflagwallbannerbuscarsbusfaces treet lampslide credit: Fei-Fei, Fergus CV Scene and context categorization outdoor city traffic ..slide credit: Fei-Fei, Fergus CV Qualitative spatial information slantedrigid moving objecthorizontalverticalslide credit: Fei-Fei, Fergus rigid moving objectnon-rigid moving object 3 2 3 2 CV 2 3 ill-posed( ) From Kentaro Toyama Almost certain to be Bill ClintonDark circular overlayGray hairNeckRight earWoman s dress suitArmani suitWhite shirtLeft eye (open)CNN caption (Washington 1995?)

4 Clintonoccluding MonicaPerson contourPerson with glasses in crowdNoseCheekMonica s mouth (smiling)LapelNecklaceRight eye (open)Dark brown hairPony tailClinton greeting LewinskyMonica LewinskyIlluminated from above6CV depthorsegmentation object pose(facing away,facing forward)actionunderstanding objectrecognition Input image From Kentaro ToyamaComputer vision vs human visionWhat we seeWhat a computer sees From Kentaro Toyama 2 (2 0~255 Challenges: viewpoint variationMichelangelo 1475-1564slide credit: Fei-Fei, Fergus & Torralba Challenges: illuminationimage credit: J. Koenderink Challenges: scaleslide credit: Fei-Fei, Fergus & Torralba 7 Challenges: deformationXu, Beihong 1943slide credit: Fei-Fei, Fergus & Torralba Challenges: occlusionMagritte, 1957 slide credit: Fei-Fei, Fergus & Torralba Challenges: object intra-class variationslide credit.)

5 Fei-Fei, Fergus & Torralba CV Image Processing Statistical Pattern Recognition Artificial Intelligence Robotics Biological vision Medical imaging Computer graphics Human-computer interaction 3D SceneLight sourceCameraPP Image PlaneImage ProcessingComputer VisionImage PlaneCGLight sourceCamera 2D 2D Computer GraphicsComputer VisionComputervisionWorld modelComputergraphicsWorld model8 1950 1960 2 J548043 Guzman `68

6 1970 1980 Ohta Kanade 78 CT An Industrial Computer vision System .. 1990 1990 VR MR Turk and Pentland 91 CV 1980 David Marr (Computation vision , 1982) [shape-from-X] 2-1/2 2 3 3 (shape-from-X)X: Shading, Texture, Contour, Stereo, Motion9 Ioannis Stamos CSc 83020 Spring 2007 Shape from ShadingThree-dimensional shape from a single is this object?

7 Does color play a role in recognition? Might this be easier to recognize from a different view?The role xof color The role x of texture Characteristic image texture can help us readily recognize objects. The role x of shape The role x of Stereodepth mapIoannis Stamos CSc 83020 Spring 2007 The role x of Motion(optical flow)Determining the movement of scene objects10 1990 3.

8 Kinect MCMC+AdaBoost Meanshift+AdaBoost 11 2D 3D 3D 2D CG Wakayama University

9 D 6612 3D 67 Applications: 3D ReconstructionPanoramic Mosaics++ .. += [Brown & Lowe 2003] 13 CG CG 2000 CG CG MatrixWarner 2 Sony Pictures Edit D CG (Virtual Reality) (Image-Base Rendering) (Virtualized Reality) VR (Augmented Reality) AR (Mixed Reality) MR Augmented RealityAndy Serkis, Gollum, Lord of the RingsApplications.

10 Special Effects Computer vision Surveillance and security Face recognition Gesture recognition Space applications Medical image analysis Autonomous vehicles Virtual reality and much more ..14 BiometricsTraffic MonitoringHuman Activity RecognitionAction recognition in videosMotion history image[Bobick & Davis, 2001] Spatial motion descriptor[Efros et al. ICCV 2003] Learning dynamic prior [Blake et al. 1998] Sign language recognition[Zisserman et al. 2009] Medical Applications skin cancerbreast cancer 15 Computer vision : vision and ModelingComputer vision : vision and ModelingAutonomous VehicleFlakey, SRI21 Reflectance Map ),,(1),,(222zyxRnnnzyyzqzxxzpyxRzzyx ( ) 16 I Iaka IikdN L ksR V n i 1nL krIr ktItka nLkd ks kr kt 192