Computer Vision CM30080

companyscourgeΤεχνίτη Νοημοσύνη και Ρομποτική

19 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

96 εμφανίσεις

Computer Vision

CM30080

Peter Hall

pmh@cs.bath.ac.uk

http://
www.cs.bath.ac.uk/~pmh/Teaching/Computer_Vision.html

Computer Vision Applications


Security


Medicine


Entertainment


Web applications


Facebook

tagging


Google image search

Computer Vision

The 3 R’s of Computer Vision


Reconstruction


Registration


Recognition

Reconstruction


Turn 2D into 3D


Shape from Shading


Shape from Motion


Shape from Texture


Shape from Stereo


Shape from N images

http://carlos
-
hernandez.org/research.html

Registration


Match things in images,
e.g.


panoramic stitching


Tracking


Recognition


What is it?


segmentation


classification

Recognition II


A video example

(thanks to Michal
Irani
, circa 2005)

Why is Vision hard?


Reconstruction


Registration


too many similar matches


obscured matches


false matches

Why is Vision Hard?

Why is Vision hard?


Recognition

Computer Vision solutions


Accounting for ambiguity, variability and error
is very hard.



Modern solutions rely heavily on


sound mathematical models


probability and statistics


Machine Learning


Extensive experimental Verification

CM30080: syllabus


Low
-
level Vision


Linear filters


Interest Points


Features



Stereo Reconstruction


Projective Geometry


Calibration


Epipolar

Constraint and the Fundamental matrix



Tracking


Kalman

filter


Condensation tracking



Segmentation


Coherence properties and spilt
-
and
-
merge


Mean shift


Berkeley watershed



Classification


Bag of Words via
Niave

Bayes



Applications


Entertainment industries


Medicine

CM30080: admin


100 hours comprising


24 lectures


11 lab sessions


self
-
study


Assessment


25
% coursework


75% examination