Object Localization with Global and Local Context Kernels

Published on 2009-12-015409 Views

Matthew B. Blaschko

Recent research has shown that the use of contextual cues significantly improves performance in sliding window type localization systems. In this work, we propose a method that incorporates both globa

BMVC 2009 - London

Related categories

Kernel Methods

Presentation

Object Localization with Global and Local Context Kernels00:00

Outline00:02

Sliding window localization00:28

Identify all Objects in an Image00:33

Sliding Window Parametrization00:48

Compatibility Function01:08

Joint kernel maps for localization02:01

Joint Kernel Maps02:02

Joint Kernel between Images and Boxes: Restriction Kerne03:08

Restriction Kernel: Examples03:52

Problem (1)04:25

Problem (2)04:34

Contextual cues04:40

Global and Local Context Kernels (1)04:44

Global and Local Context Kernels (2)05:04

Global and Local Context Kernels (3)05:24

Local Context Kernel05:32

Branch and bound localization06:28

Efficient Object Localization06:44

Branch & Bound Search - CVPR 2008, PAMI 200907:24

Sets of Rectangles08:08

Branch-Step: Splitting Sets of Boxes08:40

Bounds for Global and Local Context Kernels08:56

Results11:05

Parameter Settings11:08

Results (tabele)11:58

Conclusions (1)13:36

Conclusions (2)13:44

Conclusions (3)13:56

Thanks, Questions?14:24