How to use NimutiaeNet in image representation?

[zhaoxy92]

Hi,

Thanks for this amazing model! Could you please tell how to use this in image representation? For example, I want to compute the similarity of two fingerprints, I'm assuming that MinutiaeNet can be used to extract features from the two images so that I can use function like score(img1, img2) to find the similarity. Thank you!

[wajihullahbaig]

You don't assume, you extract features.
Once you have the minutiae, you need to perform a matching on the minutiae lists. For example, you can use the minutiae lists to generate Tico's orientation features for each fingerprint. Once you have these Orientation features you can use them to make comparisons such a Euclidean distances to find a match. Likewise MCC and minutiae triplets are ideal candidates to generate features that are minutiae-only features.
So and so forth.

Hope this helps.

[luannd]

Hi @zhaoxy92 ,

You can define any function/method to find the similarity of between 2 images based on minutiae location. For example, you can use the local patch around each minutiae, the graph structure of the whole minutiae set, the combination between location and orientation, etc. to find that score.

We will publish a paper in this fingerprint matching problem in Nov/Dec. I will notify you if you want.

[wajihullahbaig]

@luannd please let me know too when you publish the paper.

[zhaoxy92]

@wajihullahbaig Thank you! However, I'm not really a image person, do you by any chance know how exactly to map fingerprints with different number of minutiae points into vectors with a same length? Is there any implemented code to that? Thank you so much!

[zhaoxy92]

@luannd Thank you! Will you be publishing your code as well? I will ask the same question again, do you by any chance know how exactly to map fingerprints with different number of minutiae points into vectors with a same length? Is there any implemented code to that? Thank you so much!

[wajihullahbaig]

Not by chance, but I do know a few resources I can point you to.
1- You will need to fiddle with the code a bit to git it running, its in Matlab https://www.mathworks.com/matlabcentral/fileexchange/44369-fingerprint-matching-a-simple-approach

2- MCC as I mentioned above in my comments. you can find a GPU implementation here https://github.com/mvisat/mcc-cuda

Hope this helps.

[luannd]

Hi all,

Here is our NEW paper "End-to-End Latent Fingerprint Search": https://arxiv.org/pdf/1812.10213.pdf
This framework includes all steps from extraction to matching using both deep networks and parallel computing!!!

Cheers!

[wajihullahbaig]

Thanks for the paper. It's Awesome!

…

On Tue, 1 Jan 2019, 18:16 Dinh-Luan Nguyen ***@***.*** wrote: Hi all, Here is our *NEW* paper *"End-to-End Latent Fingerprint Search"*: https://arxiv.org/pdf/1812.10213.pdf This framework includes all steps from extraction to matching using both deep networks and parallel computing!!! Cheers! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#11 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACMB0D4xTfuM4szIvrMSWZEGsYRHKwROks5u-1-ngaJpZM4XBQhF> .