Skip to content

Some differences from the experimental results of Homography Estimation on the AdelaideRMF dataset. #1

Description

@hailingluo

Hi, thank you for your work. I ran the code on the Adelaide dataset for the estimation of homography matrices. However, the fitting results obtained from my own are different from yours on the paper (Fast L1-NMF for Multiple Parametric Model Estimation). I got some strange results as followed:

barrsmith (235, 6)
Preference matrix size: (235, 1425)
Running regular bi-clustering
Time: 8.2033700943
Inlier sets size: [235]
GNMI: 0.0000; Precision: nan; Recall: 0.3333; ME: 0.3021;
Running compressed bi-clustering
Time: 3.55659222603
Inlier sets size: [235]
GNMI: 0.0000; Precision: nan; Recall: 0.3333; ME: 0.3021;
Comparing both solutions:
GNMI: 0.0000; Precision: nan; Recall: nan; ME: 0.0000;

And I obtained the mean misclassification error among 19 pairs of images is up to 40%. Note that I can obtain normal fitting results on the estimation of fundamental matrices.

Looking forward to your reply. Thanks.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions