Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest...

14
Evaluation of interest points and descriptors

Transcript of Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest...

Page 1: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Evaluation of interest points and descriptors

Page 2: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Introduction

• Quantitative evaluation of interest point detectors– points / regions at the same relative location

=> repeatability rate

• Quantitative evaluation of descriptors– distinctiveness

=> detection rate with respect to false positives

Page 3: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Quantitative evaluation of detectors

• Repeatability rate : percentage of corresponding points

• Two points are corresponding if1. The location error is less than 1.5 pixel

2. The intersection error is less than 20%

homography

Page 4: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Comparison of different detectors

[Comparing and Evaluating Interest Points, Schmid, Mohr & Bauckhage, ICCV 98]

repeatability - image rotation

Page 5: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Comparison of different detectors

[Comparing and Evaluating Interest Points, Schmid, Mohr & Bauckhage, ICCV 98]

repeatability – perspective transformation

Page 6: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Harris detector + scale changes

Page 7: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Harris detector – adaptation to scale

Page 8: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Evaluation of scale invariant detectors

repeatability – scale changes

Page 9: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Evaluation of affine invariant detectors

0

40

60

70

repeatability – perspective transformation

Page 10: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Quantitative evaluation of descriptors

• Evaluation of different local features– SIFT, steerable filters, differential invariants, moment invariants,

cross-correlation

• Measure : distinctiveness– receiver operating characteristics of detection rate with respect to false positives

– detection rate = correct matches / possible matches– false positives = false matches / (database points * query points)

[A performance evaluation of local descriptors, Mikolajczyk & Schmid, CVPR’03]

Page 11: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Experimental evaluation

Page 12: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Scale change (factor 2.5)

Harris-Laplace DoG

Page 13: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Viewpoint change (60 degrees)

Harris-Affine (Harris-Laplace)

Page 14: Evaluation of interest points and descriptors. Introduction Quantitative evaluation of interest point detectors –points / regions at the same relative.

Descriptors - conclusion

• SIFT + steerable perform best

• Performance of the descriptor independent of the detector

• Errors due to imprecision in region estimation, localization