Table of contents

title

the problem 1

the problem 2

traditional approach 1

traditional approach 2

traditional approach 3

traditional approach 4

challenge: complex shape

challenge: no texture

solution: use edge cues 1

solution: use edge cues 2

classify edge pixels individually

sample result

using edge probes to classify 1

using edge probes to classify 2

using edge probes to classify 2.5

using edge probes to classify 3

using edge probes to classify 4

using edge probes to classify 5

schematic

aperture problem

aperture problem 1aa

aperture 1B

aperture 1c

aperture problem 1d

aperture 1e

aperture 1f

aperture 1g

aperture problem 2

aperture problem 3

grow the aperture 1

grow the aperture 2

grow the aperture 3

grow the aperture 4

grow the aperture 5

grow the aperture 6

grow the aperture 7

grow the aperture 8

aperture solution 1

aperture solution 2

aperture solution 3

aperture solution 4

training

a discriminative approach

experimental data : chair

experimental data : background

experimental data : composite

chair big sample result 1

chair big sample result 2

chair many results 1

chair many results 2

cart big result 1

cart big result 2

cart many result 1

cart many result 2

overall results: cart

overall results: chair

conclusion

future work

Author: Owen Carmichael

E-mail: otc@andrew.cmu.edu

Homepage: http://rope.ucdavis.edu/~owenc

Further information:
Copyright 2002 Owen Carmichael. All rights reserved.