Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Institute of Electrical Measurement and Measurement Signal Processing VISION: Axel Pinz • • • • • Paradigms Systems Algorithms Applications Evaluation A Graph-based structure ! HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 1 Institute of Electrical Measurement and Measurement Signal Processing The Graph of Vision History Systems 1978 Pattern Recognition Image Proc. 1970s Artificial Intelligence 1971 Image Understanding 1980s Cognitive Psychology 1985 ? Neurophysiology 1963 Algorithms 1962 Recognition 1970s Marr 1978 Robotics 1994 ? R+R 1990 HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Reconstruction 1980 ? Axel Pinz 9.5.2008 2 Institute of Electrical Measurement and Measurement Signal Processing Artificial Intelligence • SHRDLU [Winograd 1971] • Blocks world original screen display (MIT AI Lab) later color rendering (Univ. Utah) HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 3 Institute of Electrical Measurement and Measurement Signal Processing Blocks World – 1970s • • • • Natural language understanding Spatio-temporal reasoning LISP Visual input would be nice • Vision as an “enabling sensory module” in AI – Patrick Henry Winston “The Psychology of Computer Vision” 1975 – Marvin Minsky “A framework for representing knowledge” 1975 • Frames • Frame Representation Language FRL HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 4 Institute of Electrical Measurement and Measurement Signal Processing Blocks World – Today • Still an issue of ongoing research… • EU FP 6 project MACS • 2004-2008 Learning ! “Affordances” … HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 5 Institute of Electrical Measurement and Measurement Signal Processing AI Methods for Vision – 1980s • “Image Understanding” • Expert systems – SIGMA Aerial Image Understanding – Matsuyama, Hwang, … • Reasoning, Blackboard system – VISIONS, Draper, OHM, … • Schema learning – SLS, Draper,… HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 6 Institute of Electrical Measurement and Measurement Signal Processing SIGMA [Matsuyama] HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 7 Institute of Electrical Measurement and Measurement Signal Processing VISIONS [Hanson+Riseman], SLS [Draper] A Massachusetts “road”-scene HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 8 Institute of Electrical Measurement and Measurement Signal Processing “Image Understanding” – 1980s • Integrating bottom-up and top-down processes combinatorial explosion • “knowledge engineers”, hand-crafted knowledge-base learning is a must HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 9 Institute of Electrical Measurement and Measurement Signal Processing Systems 1990 – 200+ • “Computer Vision Systems” – [Hanson+Riseman 1978] • UMass VISIONS KBVision – AAI Amerinex Artificial intelligence – AAI Amerinex Applied Imaging • Grimson Cognex – PatMax • IUE – TargetJr Joe Mundy • Matlab • OpenCV • toolboxes, libraries HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 10 Institute of Electrical Measurement and Measurement Signal Processing Robotics • Brooks: “building brains for bodies” – Autonomous Robots 1:7-25, 1994 • Learning! • Interaction! • Reward by survival…! • Active, purposive, qualitative – Bajcsy 1988 – Aloimonos 1989 HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 11 Institute of Electrical Measurement and Measurement Signal Processing Babybot – Lira Lab [Sandini] learning to push learnt Babybot’s view HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 12 Institute of Electrical Measurement and Measurement Signal Processing Neurophysiology • • • • Retina visual cortex Layered representation in visual cortex Receptive fields Hubel+Wiesel – – – – – Gradients Oriented “edgels” At various scales Bottom-up grouping recognition Visual pathway stereo, reconstruction HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 13 Institute of Electrical Measurement and Measurement Signal Processing Visual Pathway [Hubel+Wiesel] HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 14 Institute of Electrical Measurement and Measurement Signal Processing Receptive Fields [Hubel+Wiesel] on-off simple complex HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 15 Institute of Electrical Measurement and Measurement Signal Processing Pattern Recognition + Image Processing – PRIP • Structural [Pavlidis 1972, Fu 1982] • Statistical [Fukunaga 1990] • Representation (regular, irregular, scale,…) pyramids, scale space, graphs,… • Feature sets, feature selection, classifiers, learning,… • 2D (discrete) signal processing HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 16 Institute of Electrical Measurement and Measurement Signal Processing Maximum Likelihood Classification HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 17 Institute of Electrical Measurement and Measurement Signal Processing Scale Space [Lindeberg] Space + time [Laptev + Lindeberg] HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 18 Institute of Electrical Measurement and Measurement Signal Processing Cognitive Psychology • Perceptual grouping – Lowe 1985 – Sarkar 1994 – LLVE Matsuyama • Qualitative volumetric models (Geons) – Biederman 1985, Bergevin+Levine 1988 – Dickinson 1992: Aspect Hierarchy • A renaissance in today’s shape models for category detection… HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 19 Institute of Electrical Measurement and Measurement Signal Processing My Favorite Example [Lowe] Very hard !!! HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 20 Institute of Electrical Measurement and Measurement Signal Processing My Favorite Example [Lowe] Very easy !!! HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 21 Institute of Electrical Measurement and Measurement Signal Processing Geons [Biederman], Aspect Hierarchy [Dickinson] HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 22 Institute of Electrical Measurement and Measurement Signal Processing Geons [Biederman], Aspect Hierarchy [Dickinson] HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 23 Institute of Electrical Measurement and Measurement Signal Processing The Marr Paradigm – 1978-? • Systems approach to vision computational model – Algorithm, data, hardware implementation – Low level modules • Representational levels – Primal sketch saliency – 2-1/2 D reconstruction ! – 3D object model is it really required? • • • • “reconstruction school” “visual modules”, “shape from X” Zerroug 1994 Structure + Motion 200x HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 24 Institute of Electrical Measurement and Measurement Signal Processing Marr (~1978) – “primal sketch” “saliency” ! “Harris” corners HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 25 Institute of Electrical Measurement and Measurement Signal Processing Marr (~1978) – “2-1/2-D sketch”, “generalized cone” HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 26 Institute of Electrical Measurement and Measurement Signal Processing Marr (~1978) – 3D hierarchical models HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 27 Institute of Electrical Measurement and Measurement Signal Processing Marr ~1978, Aloimonos ~1988 “robustness” HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 28 Institute of Electrical Measurement and Measurement Signal Processing Zerroug 1994 HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 29 Institute of Electrical Measurement and Measurement Signal Processing Algorithms • • • • • • • • • • • • • Convolution kernels Mathematical morphology structuring elements Serra Fourier transform spectrum, phase, descriptors, Gabor filter bank Hough transform voting space PCA Pyramids (regular, irregular) graphs Scale space theory in CV Lindeberg Algebraic projective geometry Hartley, Zisserman Color (good chapters in Forsyth+Ponce, Burger+Burge) Texture invariant moments (Hu … Haralick … VanGool) Interest point detection (Schmid, …) affine covariance Pattern Classification Duda+Hart Machine Learning Hastie et al. HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 30 Institute of Electrical Measurement and Measurement Signal Processing Reconstruction • • • • Projective geometry Photogrammetry Essential, fundamental, multiview, … Calibrated, uncalibrated, autocalibration, calibration-free,… • Structure and Motion HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 31 Institute of Electrical Measurement and Measurement Signal Processing Structure and motion [Schweighofer 2008] HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 32 Institute of Electrical Measurement and Measurement Signal Processing Recognition • Specific objects – PCA [Murase+Nayar 95], Eigenfaces [Pentland 93] • Categories – Generative [Fergus 2003], discriminative [Opelt 2004] • Recognition vs. localization • Features (descriptors), grouping – Shape, texture, color, proximity, similarity,… • Saliency (detectors) • Active – Motion, space+time … HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 33 Institute of Electrical Measurement and Measurement Signal Processing Shape-Based Category Detection [Opelt 2006] HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 34 Institute of Electrical Measurement and Measurement Signal Processing “R + R”: Confluence of Recognition and Reconstruction • “recognition and reconstruction schools will merge” [Aloimonos+Shulman 1989] • “Marr paradigm – use of 2-1/2D complicates the problem” [Medioni et al. 2000] tensor voting • Spatio-temporal reasoning and representation HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Axel Pinz 9.5.2008 35 Institute of Electrical Measurement and Measurement Signal Processing The Graph of Vision History Systems 1990s Pattern Recognition Image Proc. 1970s Artificial Intelligence 1971 Image Understanding 1980s Cognitive Psychology 1985 ? Neurophysiology 1963 Algorithms 1962 Recognition 1970s Marr 1978 Robotics 1994 ? R+R 1990 HistoryHorst of Computer Vision Paradigms-Systems-Algorithms-Applications-Evaluation Professor Cerjak, 19.12.2005 Reconstruction 1980 ? Axel Pinz 9.5.2008 36