Matching questions

General information about Neurotechnology and its software

Re: Advice on fingerprints falsely rejected

Postby monkeyhandz » Nov 29, 2013 10:30

Hi

As requested I have sent an email to support regarding this but have had no reply or acknowledgement so far.

I'm still abit confused by the documentation or the answer, it isn't clear to me whether I should increase the matchingThreshold on the NMatcher to reduce FFR's or vice versa. Want to be clear I'm talking about attribute on the NMatcher, not the result score returned by the matcher (If the score is greater than zero we accept it as a match).

Also what I'm finding the FFR rate reduces if I set to NMatacher.matchingThreshold to 6, but I still get FFR's. So I tried to increase the NMatcher.matchingThreshold to 24 but this only increases the number of FFR's again!

Kind regards
David Smith
monkeyhandz
 
Posts: 16
Joined: Jul 21, 2013 19:24

Re: Advice on fingerprints falsely rejected

Postby vaidasz » Nov 29, 2013 14:03

Hello David,

monkeyhandz wrote:As requested I have sent an email to support regarding this but have had no reply or acknowledgement so far.

The e-mail reply could take up to 2 business days. We also have different time zone that could cause time for e-mail response.
The e-mail has been answered to you about half an hour ago.

monkeyhandz wrote:I'm still abit confused by the documentation or the answer, it isn't clear to me whether I should increase the matchingThreshold on the NMatcher to reduce FFR's or vice versa. Want to be clear I'm talking about attribute on the NMatcher, not the result score returned by the matcher (If the score is greater than zero we accept it as a match).
Also what I'm finding the FFR rate reduces if I set to NMatacher.matchingThreshold to 6, but I still get FFR's. So I tried to increase the NMatcher.matchingThreshold to 24 but this only increases the number of FFR's again!

If the score matched between templates is lower than the matching threshold set, 0 is returned. Otherwise, the score is returned.
For example:
• If you set matching threshold to 0, all the templates that matches with score above 0 will be returned.
• If you set matching threshold to 24, all the templates that matches with score 24 or more will return the score that it matched with.
• If you set matching threshold to 48, all the templates that matches with score 48 or more will return the score that it matched with.
• And so on.
The higher is the matching score, the higher is the probability that the fingerprint is of the same person.
The higher is the matching threshold set, the lower is the FAR (false acceptance rate) but the higher is the FRR (false rejection rate).

There is also a formula provided in the documentation how to calculate false acceptance probability of your database:
(1-(1-FAR/100)^N)*100, where N - DB size
For example:
If FAR=0.001% then probability that false acceptance situation will occur during 1:N identification (where N=10 000) is
1-(1-0.00001)^10000=9.52%.
If FAR=0.0001% then probability that false acceptance situation will occur during 1:N identification (where N=10 000) is
1-(1-0.000001)^10000=1.00%.


You have sent the templates. They seem to be partially scanned. For example, bottom of the fingerprint scanned, template extracted and enrolled to the database. Then the top of the fingerprint is scanned, template extracted and sent for matching. As templates contains features of different fingerprint's parts, they do not have many features set that could be matched. This is why low score is returned (like 6 or 10). At least not it looks like you are trying to match different parts of the fingerprint.
It is needed to get the fingerprint images to be able to tell more.
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

Re: Matching questions

Postby jorgevaron » Jan 07, 2014 14:39

hello,

We acquired VeriFinger Extended 6.4 SDK in 2012:

Fingerprint Extractor, Fingerprint Matcher, Fingerprint Client and Fingerprint Client Concurrent licenses.

How can we implement a way to locate a person in an Oracle database, knowing only their enrolled fingerprint ?

thanks in advance.
jorgevaron
 
Posts: 26
Joined: Dec 14, 2011 19:05

Re: Matching questions

Postby Martynas » Jan 08, 2014 09:33

jorgevaron wrote:hello,

We acquired VeriFinger Extended 6.4 SDK in 2012:

Fingerprint Extractor, Fingerprint Matcher, Fingerprint Client and Fingerprint Client Concurrent licenses.

How can we implement a way to locate a person in an Oracle database, knowing only their enrolled fingerprint ?

thanks in advance.

Hello,

as you wrote, you are using VeriFinger Extended 6.4. The current version is 6.5 and I would recommend to upgrade to latest one.
In any case as you are using the Extended Version of the VeriFinger SDK, then you have a possibility to use the NServer, a ready-to-use matching server. NServer supports the Oracle database. So you can configure the NServer to use the Oracle table with enrolled fingerprint templates (extracted from fingerprint images). Once it is done and the NServer is running you can run the client application, which will send a scanned fingerprint to NServer for the identification. After that, if the finger exists in the database, the NServer will return the id of the matched template. How to send and receive tasks/results to/from the NServer is shown in AbisSample and tutorials, included in the SDK.
AbisSample and NServer configuration is described in the documentation of the SDK.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Matcher params in C++

Postby nurzico » May 18, 2014 10:52

I create HNMatcher by NMCreate(&matcher) by this method.
I need to set Finger matching speed, Finger min matched count and matching threshold.
I tried many ways. but i failed.
Help me please.

Admin: merged to "Matching questions"
nurzico
 
Posts: 25
Joined: Sep 25, 2013 07:36

Re: Matcher params in C++

Postby Martynas » May 19, 2014 08:51

nurzico wrote:I create HNMatcher by NMCreate(&matcher) by this method.
I need to set Finger matching speed, Finger min matched count and matching threshold.
I tried many ways. but i failed.
Help me please.

Hi,

What is the revision number of the SDK you are using? It can be found in the SDKRevision.txt file of the downloaded SDK.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Matching score of similarity meaning

Postby guilespi » May 19, 2014 17:00

Hi

I would like to know what the similarity score means.

I obviously assume higher is better, but an actual probabilistic score or p-value would be what I need in order to understand what am I looking at.

I guess it has some relation with the thresholds configuration meaning a score 96 is 0.000001% of uncertainty?

What's the rule to derive probability from the following score list then? 64756? I've seen scores in the hundreds of thousands.

Code: Select all
({:id "17_5.bmp.dat",
  :score 64576,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_12.bmp.dat",
  :score 1883,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_6.bmp.dat",
  :score 895,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_3.bmp.dat",
  :score 237,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_1.bmp.dat",
  :score 223,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_2.bmp.dat",
  :score 140,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_7.bmp.dat",
  :score 137,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_8.bmp.dat",
  :score 126,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_9.bmp.dat",
  :score 96,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil})


I would like to know the right "formula" to go from similarity score to probabilistic score.

Thanks
G.-

Admin: merged to "Matching questions"
guilespi
 
Posts: 11
Joined: Apr 29, 2014 02:56

Re: Matching score of similarity meaning

Postby Martynas » May 20, 2014 07:19

guilespi wrote:Hi

I would like to know what the similarity score means.

I obviously assume higher is better, but an actual probabilistic score or p-value would be what I need in order to understand what am I looking at.

I guess it has some relation with the thresholds configuration meaning a score 96 is 0.000001% of uncertainty?

What's the rule to derive probability from the following score list then? 64756? I've seen scores in the hundreds of thousands.

Code: Select all
({:id "17_5.bmp.dat",
  :score 64576,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_12.bmp.dat",
  :score 1883,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_6.bmp.dat",
  :score 895,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_3.bmp.dat",
  :score 237,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_1.bmp.dat",
  :score 223,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_2.bmp.dat",
  :score 140,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_7.bmp.dat",
  :score 137,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_8.bmp.dat",
  :score 126,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil}
 {:id "17_9.bmp.dat",
  :score 96,
  :details #<NMatchingDetails NMatchingDetails>,
  :subject nil})


I would like to know the right "formula" to go from similarity score to probabilistic score.

Thanks
G.-

Hello,

Matching score represents the probability of the false acceptance. The higher the matching score is returned the smaller the probability of the false acceptance. You can set the matching threshold according to your project requirements or the field tests results. When the matching is done, the matching threshold is used in this way: if the matching score is equal or higher than the set matching threshold, then the matching score is returned as result. If during the matching the matching score is lower than the matching threshold, than the zero value is returned.
More information can be found in documentation of the SDK, section "6.1.1 Matching Threshold and FAR/FRR", where you will find formula for converting score to FAR too.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Re: Matching score of similarity meaning

Postby guilespi » May 20, 2014 14:44

Hi Martyna, thanks for your answer.

Matching score obviously is NOT probability of false acceptance, since a probability value is by definition between 0 and 1, and similarity scores here are taking values like 156723, 19 and 0.

I've read the documentation before asking the question and did ask anyway because it's not clear.

Doc says:

Code: Select all
Threshold = -12 * log10(FAR)


That means if I have a "similarity score" of 64576 the FAR for that value is: 10 ^ ( score / -12) ? meaning 10 ^ (-5381) which is almost 0?

And why are you presenting a "similarity score" which is not the real probability score but a transformed value?

Thanks
G.-
guilespi
 
Posts: 11
Joined: Apr 29, 2014 02:56

Re: Matching score of similarity meaning

Postby Martynas » May 21, 2014 07:09

Hi,

guilespi wrote:Matching score obviously is NOT probability of false acceptance, since a probability value is by definition between 0 and 1, and similarity scores here are taking values like 156723, 19 and 0.


It is a FAR, but expressed in integer values. Therefore we provide the formula in documentation, which allows to convert it to % expression of the FAR.

guilespi wrote:That means if I have a "similarity score" of 64576 the FAR for that value is: 10 ^ ( score / -12) ? meaning 10 ^ (-5381) which is almost 0?


yes, this is correct, with such matching score the probability that the false acceptance happened is almost zero.

guilespi wrote:And why are you presenting a "similarity score" which is not the real probability score but a transformed value?


It is our choice, as it allows us to represent it more user friendly.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Re: Matcher params in C++

Postby nurzico » May 21, 2014 09:10

SDK revision number is 88532
nurzico
 
Posts: 25
Joined: Sep 25, 2013 07:36

Re: Matcher params in C++

Postby Martynas » May 21, 2014 10:35

nurzico wrote:SDK revision number is 88532

please use NObjectSetParameter function to set parameters. The list of parameters which is possible to set is provided int he documentation of the SDK, section "8.1.2.26.3 Macros".
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Re: Matcher params in C++

Postby nurzico » May 21, 2014 19:53

I have used NObjectSetParameter method like extractor to set parameter from sample code. But i have got error for invalid parameter. I couldn't find any sample code for matcher parameter. I tried in many ways. But always i have got error.
Can u please give me a sample code for matcher? Thanks
nurzico
 
Posts: 25
Joined: Sep 25, 2013 07:36

Re: Matcher params in C++

Postby Martynas » May 22, 2014 07:11

Hi,

nurzico wrote:I have used NObjectSetParameter method like extractor to set parameter from sample code. But i have got error for invalid parameter. I couldn't find any sample code for matcher parameter. I tried in many ways. But always i have got error.


Could you please provide detailled information how it was used and what error did you get?

nurzico wrote:Can u please give me a sample code for matcher?


Code: Select all
   result = NMCreate(&matcher);
   if (NFailed(result))
   {
      PrintErrorMsg(N_T("failed to create matcher, error %d\n"), result);
      goto FINALLY;
   }

   speed = nmsHigh;

   result = NObjectSetParameter(matcher, NMP_FINGERS_MATCHING_SPEED, &speed);
   if (NFailed(result))
   {
      PrintErrorMsg(N_T("failed to set matching speed, error %d\n"), result);
      goto FINALLY;
   }

   minCount = 7;

   result = NObjectSetParameter(matcher, NMP_FINGERS_MIN_MATCHED_COUNT, &minCount);
   if (NFailed(result))
   {
      PrintErrorMsg(N_T("failed to set min count, error %d\n"), result);
      goto FINALLY;
   }

   threshold = 92;

   result = NObjectSetParameter(matcher, NMP_MATCHING_THRESHOLD, &threshold);
   if (NFailed(result))
   {
      PrintErrorMsg(N_T("failed to set threshold, error %d\n"), result);
      goto FINALLY;
   }
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

How to Create NImage from binary file?

Postby magictw » Jun 18, 2014 06:36

HI

I use "Neurotec Biometric 5.0 SDK " to implement a face recognition program.

I Extract face from an image and use NObjectSaveToMemoryN function to save the features.

Now , I have a problem to load the feature files for doing face recognition.

I tried to use NImageCreateFromFileEx to load the features , but it doesn't work.

How can I load these files?

Admin: merged to "Matching questions"
magictw
 
Posts: 6
Joined: Jun 18, 2014 06:27

Re: How to Create NImage from binary file?

Postby Martynas » Jun 18, 2014 07:01

magictw wrote:HI

I use "Neurotec Biometric 5.0 SDK " to implement a face recognition program.

I Extract face from an image and use NObjectSaveToMemoryN function to save the features.

Now , I have a problem to load the feature files for doing face recognition.

I tried to use NImageCreateFromFileEx to load the features , but it doesn't work.

How can I load these files?

Hello,

as you wrote you have saved extracted features. Features does not include the image. So it is not possible to use the function NImageCreateFromFileEx to load the extracted features. If you need to load features from the file, then you need to read it as byte array and use it for matching. You can take a look at tutorials "Verify" or "Identify", which shows how to load the binary file.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Re: Face Identification and Add to Server Database

Postby ceciliafava » Aug 12, 2014 13:28

I am using MegaMatcher SDK but I don't have any cluster server. I need compare a template with a templates stored in a database.
The tutorials that are using for .NET system web-based , use
Code: Select all
 NBiometricTask enrollTask = biometricClient.CreateTask(NBiometricOperations.Enroll, null);


a enrollment task before using Identify function to match candidates for find similarities.
Is it neccesary to do the enrollment in memory?
Thanks,
ceciliafava
 
Posts: 6
Joined: Jun 18, 2014 18:39

Re: Face Identification and Add to Server Database

Postby Martynas » Aug 13, 2014 06:37

ceciliafava wrote:I am using MegaMatcher SDK but I don't have any cluster server. I need compare a template with a templates stored in a database.
The tutorials that are using for .NET system web-based , use
Code: Select all
 NBiometricTask enrollTask = biometricClient.CreateTask(NBiometricOperations.Enroll, null);


a enrollment task before using Identify function to match candidates for find similarities.
Is it neccesary to do the enrollment in memory?
Thanks,

Hello,

if you do not use the matching server, then you need to do the matching locally and the templates used for matching should be in memory. In case if you are want to perform the Identification (1:N) matching, then it is needed to read all templates and insert them into engine. How to verify or identify templates is shown in the tutorials provided in the SDK.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Re: Matching questions

Postby ceciliafava » Sep 05, 2014 12:20

I am using NServer , MegaMatcher Extended 5 SDK.
I am testing matching faces con NServer. I did 7000 face extractions , in SQL Server Database, table usertbl.
In the Nserver.conf

Code: Select all
 SQL additional data specification
Server.SqlColumnMetaData =
# SQL data query, data will be splitted to all nodes
Server.SqlQueryPattern     = select dbid, template, hash_name from "usertbl"
# SQL select statement to fetch hash from sql table (required for spare nodes)
Server.SqlHashQueryPattern    = select dbid, template, hash_name from "usertbl"  where hash_name=@hash@
# Data update query, data will be updated for all nodes
Server.SqlUpdateQueryPattern  = select dbid, template, hash_name from "usertbl" where dbid = @id@
# Insert query pattern
Server.SqlInsertQueryExPattern = insert into "usertbl" (dbid, template, hash_name) values (@[id]@, @[template]@, @[hash]@)
# Query delete template
Server.SqlDeleteQueryPattern  = delete from "usertbl" where dbid=@1@
# Data query column name for splitting data to nodes
Server.SqlUidColumnName           = dbid
# Data query column name for extracted fingerprint data
Server.SqlTemplateColumnName     = template
# Column name for hash in SQL table (required for spare nodes)
Server.SqlHashColumnName         = hash_name

# Maximum running task count
Server.MaxTaskCount = 1000

# Database for integrated node
Node.SqlDriverName     = Integrated


And when I have sent the matching it does against memory, I read documentations, sampes, tutorials.
The matching:

Code: Select all
 private Communication commClient = new Communication(server, 25452);

        private const string query = "SELECT node_id, dbid FROM node_tbl ";
       SendRequest(commClient, templateResultado, query, out taskId);

        resultados = WaitForResult(commClient, taskId);



Only find the template that I am comparing!
In fact, I sent a template and only compare with the template that it had extracteed! I seems doing the comparation in memory!
Thanks
ceciliafava
 
Posts: 6
Joined: Jun 18, 2014 18:39

NServer for verify 1:1

Postby dicaalba » Sep 05, 2014 15:35

Hi! I'm revising the codes from the Samples/Biometrics folder and any code contains a method for verification 1 to 1 and any of these are using NServer for this, these examples only described verification in local. is it possible work with NServer and Verification 1:1 ?

Product revision number: 120480

Admin: merged to "Matching questions"
dicaalba
 
Posts: 10
Joined: Jun 06, 2013 21:13

Re: Matching questions

Postby vaidasz » Sep 05, 2014 15:50

ceciliafava wrote:I am using NServer , MegaMatcher Extended 5 SDK.
I am testing matching faces con NServer. I did 7000 face extractions , in SQL Server Database, table usertbl.
In the Nserver.conf

Code: Select all
 SQL additional data specification
Server.SqlColumnMetaData =
# SQL data query, data will be splitted to all nodes
Server.SqlQueryPattern     = select dbid, template, hash_name from "usertbl"
# SQL select statement to fetch hash from sql table (required for spare nodes)
Server.SqlHashQueryPattern    = select dbid, template, hash_name from "usertbl"  where hash_name=@hash@
# Data update query, data will be updated for all nodes
Server.SqlUpdateQueryPattern  = select dbid, template, hash_name from "usertbl" where dbid = @id@
# Insert query pattern
Server.SqlInsertQueryExPattern = insert into "usertbl" (dbid, template, hash_name) values (@[id]@, @[template]@, @[hash]@)
# Query delete template
Server.SqlDeleteQueryPattern  = delete from "usertbl" where dbid=@1@
# Data query column name for splitting data to nodes
Server.SqlUidColumnName           = dbid
# Data query column name for extracted fingerprint data
Server.SqlTemplateColumnName     = template
# Column name for hash in SQL table (required for spare nodes)
Server.SqlHashColumnName         = hash_name

# Maximum running task count
Server.MaxTaskCount = 1000

# Database for integrated node
Node.SqlDriverName     = Integrated


And when I have sent the matching it does against memory, I read documentations, sampes, tutorials.
The matching:

Code: Select all
 private Communication commClient = new Communication(server, 25452);

        private const string query = "SELECT node_id, dbid FROM node_tbl ";
       SendRequest(commClient, templateResultado, query, out taskId);

        resultados = WaitForResult(commClient, taskId);



Only find the template that I am comparing!
In fact, I sent a template and only compare with the template that it had extracteed! I seems doing the comparation in memory!
Thanks


Hello,

Could you please be more specific and explain in more details what is the issue? I do not understand it.
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

Re: NServer for verify 1:1

Postby vaidasz » Sep 05, 2014 15:52

dicaalba wrote:Hi! I'm revising the codes from the Samples/Biometrics folder and any code contains a method for verification 1 to 1 and any of these are using NServer for this, these examples only described verification in local. is it possible work with NServer and Verification 1:1 ?

Product revision number: 120480


Hello,

Yes, it is possible. Just add "WHERE dbid=ID_YOU_WANT_TO_VERIFY_WITH" at the end of matching task.
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

Re: Matching questions

Postby ceciliafava » Sep 17, 2014 11:47

I am using NServer and I would like to set the THRESHOLD
MatchingParameters.AddParameter Method : I don't know the parameters values I have to use.

Any Help?
Thanks
ceciliafava
 
Posts: 6
Joined: Jun 18, 2014 18:39

Re: Matching questions

Postby vaidasz » Sep 17, 2014 15:21

ceciliafava wrote:I am using NServer and I would like to set the THRESHOLD
MatchingParameters.AddParameter Method : I don't know the parameters values I have to use.

Any Help?
Thanks


Hello,

To set matching threshold use:
Code: Select all
MatchingParameters parameters = new MatchingParameters();
parameters.AddParameter( NMatcher.PartNone, NMatcher.ParameterMatchingThreshold, threshold);
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

Re: Matching questions

Postby ceciliafava » Sep 19, 2014 17:00

I have tried NMatcher.PartNone , I am using NServer inside an application web based over .NET
The .NET said that the NMatcher is deprecated and PartNone doesn't exist.
I need to set the parameters, one of them is NMP_MATCHING_THRESHOLD to match faces and fingers

Regards,
ceciliafava
 
Posts: 6
Joined: Jun 18, 2014 18:39

Re: Matching questions

Postby vaidasz » Sep 22, 2014 12:14

ceciliafava wrote:I have tried NMatcher.PartNone , I am using NServer inside an application web based over .NET
The .NET said that the NMatcher is deprecated and PartNone doesn't exist.
I need to set the parameters, one of them is NMP_MATCHING_THRESHOLD to match faces and fingers

Regards,


Hello,

I assume you are talking about Neurotec Biometric 5.0 SDK. In such case, please ignore that NMathcher is deprecated.
To set matching threshold use:
Code: Select all
parameters.AddParameter(NMatcher.ParameterMatchingThreshold, threshold);
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

Relation between match threshold and min match count thresho

Postby nurzico » Sep 27, 2014 20:25

What is the relation between match threshold and min match count threshold in respect to min match count?
I am using combined fingerprint templates to match.
In example, match threshold 96, min match count 2 and min match count threshold 96
Matching two combined templates each having 4 fingers.
Can you please explain what is the scenario of this case for matching? I mean how min match count threshold relating with match threshold and match count for deciding whether the pair is matched or not?
I am using SDK version 4.3

Admin: merged to "Matching questions"
nurzico
 
Posts: 25
Joined: Sep 25, 2013 07:36

Score in percentage

Postby nurzico » Sep 27, 2014 20:41

I have seen same type of question's answer.
I know there is no limit of matching score.
But is there any way to figure it out the fingerprint similarity score in percentage or any classifications?
I am using combined fingerprint templates to match.
My SDK version is 4.3

Sorry to ask that question again.

Admin: merged to "Matching questions"
nurzico
 
Posts: 25
Joined: Sep 25, 2013 07:36

Re: Score in percentage

Postby Martynas » Oct 06, 2014 12:08

nurzico wrote:I have seen same type of question's answer.
I know there is no limit of matching score.
But is there any way to figure it out the fingerprint similarity score in percentage or any classifications?
I am using combined fingerprint templates to match.
My SDK version is 4.3

Sorry to ask that question again.

Hello,

No, it is not possible to convert the similarity score into percentage.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Re: Relation between match threshold and min match count thr

Postby Martynas » Oct 06, 2014 12:14

nurzico wrote:What is the relation between match threshold and min match count threshold in respect to min match count?
I am using combined fingerprint templates to match.
In example, match threshold 96, min match count 2 and min match count threshold 96
Matching two combined templates each having 4 fingers.
Can you please explain what is the scenario of this case for matching? I mean how min match count threshold relating with match threshold and match count for deciding whether the pair is matched or not?
I am using SDK version 4.3

Hello,

i.e. you have 4 fingerprint template. i.e. If the Minimal Match count is set to two, then during matching initially only 2 fingers from 4 fingers template will be used for matching and if their score will be higher then the set Minimal Match Threshold, then other two fingers will be matched too. Otherwise, if the score will be lower than the set Minimal Match Threshold then other two fingers are not matched.
This helps to speed up the matching, but could result in lower accuracy.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Re: Relation between match threshold and min match count thr

Postby vaidasz » Oct 06, 2014 12:44

nurzico wrote:What is the relation between match threshold and min match count threshold in respect to min match count?
I am using combined fingerprint templates to match.
In example, match threshold 96, min match count 2 and min match count threshold 96
Matching two combined templates each having 4 fingers.
Can you please explain what is the scenario of this case for matching? I mean how min match count threshold relating with match threshold and match count for deciding whether the pair is matched or not?
I am using SDK version 4.3


Hello,

FingersMinMatchedCount and FingersMinMatchedCountThreshold are used for customized matching. Please refer to the documentation of the SDK, section “Cuztomized matching”.
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

Re: Score in percentage

Postby vaidasz » Oct 06, 2014 12:44

nurzico wrote:I have seen same type of question's answer.
I know there is no limit of matching score.
But is there any way to figure it out the fingerprint similarity score in percentage or any classifications?
I am using combined fingerprint templates to match.
My SDK version is 4.3

Sorry to ask that question again.


Hello,

No, there is no possibility to convert matching threshold to percentage value.
As you may know the matching threshold represents the probability that the false acceptance happened. The higher the returned matching score, the lower the probability that the false acceptance happened.
Please read documentation, section “Matching Threshold and FAR/FRR” to understand better.
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

How can i user face fast matcher [Matching]

Postby Paulo » Dec 05, 2014 11:15

What component i need to use to get face fast matcher

Admin: merged to "Matching questions"
Paulo
 
Posts: 3
Joined: Dec 04, 2014 19:51

Re: How can i user face fast matcher

Postby Martynas » Dec 08, 2014 09:30

Paulo wrote:What component i need to use to get face fast matcher

Hello,

you need to have Fast Face Matcher license activated on that machine.
In your code you need to obtain license for "Biometrics.FaceMatchingFast" component.
Also you need to set FacesMatchingSpeed parameter in the NBiometricClient to High.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Re: Matching questions

Postby Wisarut » Jan 29, 2015 16:05

Now, I have try using AbisSample (AbisSampleWX.exe) for face detection but still don"t understand how to enroll the facial image files taken from image file for officer ID cards to train AbisSample to learn about the faces of 18 men and women in my office and then detect the facial image files taken from video clips which has been kept in various directories. AbisSample does not as straightforward as FacialSample program (FacesSampleWX.exe) at all. If you can explain how to use AbisSample for proper facial detection, I would be appreciated.

Furthermore, the suggestion on how to modify source codes of FacialSample program to allow straightforward facial detection with function to readjust (lower) Quality Threshold to make FacialSample program to learn about the facial information taken from ID cards to detect facial image files taken from video clips to identify whether the facial image files taken from video clips is one of 18 men and women in my office. If FacesSample program is written in C or C++, I would be greatly appreciated.
Wisarut
 
Posts: 28
Joined: Jan 16, 2015 10:07

Re: Matching questions

Postby Martynas » Feb 02, 2015 08:21

Wisarut wrote:Now, I have try using AbisSample (AbisSampleWX.exe) for face detection but still don"t understand how to enroll the facial image files taken from image file for officer ID cards to train AbisSample to learn about the faces of 18 men and women in my office and then detect the facial image files taken from video clips which has been kept in various directories. AbisSample does not as straightforward as FacialSample program (FacesSampleWX.exe) at all. If you can explain how to use AbisSample for proper facial detection, I would be appreciated.

Furthermore, the suggestion on how to modify source codes of FacialSample program to allow straightforward facial detection with function to readjust (lower) Quality Threshold to make FacialSample program to learn about the facial information taken from ID cards to detect facial image files taken from video clips to identify whether the facial image files taken from video clips is one of 18 men and women in my office. If FacesSample program is written in C or C++, I would be greatly appreciated.

Hello,

please write an email to support@neurotechnology.com with the issue description.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Method Verify NBiometricEngine

Postby Jorgtdna » Jul 23, 2015 13:04

Method NBiometricClient.Verify It just work in server, and after some time, the performance is low, takes longer to process. It starts fast and after turns slow. Why? I"m doing multiples verification within loop.

Which is the way to do the multiple verification without using a lot of memory? the memory machine is increasing quite.

Code: Select all
                var connection = new NClusterBiometricConnection { Host = IP_SERVIDOR, AdminPort = PUERTO_SERVIDOR };
                localBiometric.RemoteConnections.Add(connection);
                localBiometric.MatchingWithDetails = true;
                localBiometric.MatchingThreshold = UMBRAL;
                localBiometric.FingersMatchingSpeed = NMatchingSpeed.Medium;
                candidato.QueryString = String.Format("Id = {0}", cedula);
                sujeto2 = db.getTemplatePersona(cedula);
                sujeto2.QueryString = String.Format("Id = {0}", cedula);
                IAsyncResult result = localBiometric.BeginVerify(candidato, sujeto2, null, null);
                NBiometricStatus status = localBiometric.EndVerify(result);

                if (status == NBiometricStatus.Ok || status == NBiometricStatus.MatchNotFound)


Admin: merged to "Matching questions"
Jorgtdna
 
Posts: 3
Joined: Jul 22, 2015 05:04

Method Verify or Identify

Postby Jorgtdna » Jul 23, 2015 13:07

Which Method is better to use, Verify or Indentify? NBiometricEngine class methods.

Identify is runs faster, Verify is low. why?
Jorgtdna
 
Posts: 3
Joined: Jul 22, 2015 05:04

Re: Method Verify NBiometricEngine

Postby MartynasV » Jul 24, 2015 09:16

Hello Jorgtdna,
Jorgtdna wrote:Method NBiometricClient.Verify It just work in server, and after some time, the performance is low, takes longer to process. It starts fast and after turns slow. Why? I"m doing multiples verification within loop.

Which is the way to do the multiple verification without using a lot of memory? the memory machine is increasing quite.

Code: Select all
                var connection = new NClusterBiometricConnection { Host = IP_SERVIDOR, AdminPort = PUERTO_SERVIDOR };
                localBiometric.RemoteConnections.Add(connection);
                localBiometric.MatchingWithDetails = true;
                localBiometric.MatchingThreshold = UMBRAL;[quote="Jorgtdna"]Which Method is better to use, Verify or Indentify? NBiometricEngine class methods.

Identify is runs faster, Verify is low. why?[/quote]
                localBiometric.FingersMatchingSpeed = NMatchingSpeed.Medium;
                candidato.QueryString = String.Format("Id = {0}", cedula);
                sujeto2 = db.getTemplatePersona(cedula);
                sujeto2.QueryString = String.Format("Id = {0}", cedula);
                IAsyncResult result = localBiometric.BeginVerify(candidato, sujeto2, null, null);
                NBiometricStatus status = localBiometric.EndVerify(result);

                if (status == NBiometricStatus.Ok || status == NBiometricStatus.MatchNotFound)

Dispose all disposable objects, when you not using them anymore.

Jorgtdna wrote:Which Method is better to use, Verify or Indentify? NBiometricEngine class methods.

Identify is runs faster, Verify is low. why?

Verification is for 1:1 matching. Identification - 1:many matching.
If you are using Verify to do 1:many matching, it will always perform few times slower than Identify, because identification is optimized to do 1:N matching and verification is not. Use identification to do 1:N matching.
Martynas V
Neurotechnology
MartynasV
Neurotechnology
 
Posts: 258
Joined: Feb 27, 2015 11:24
Location: Vilnius, Lithuania

Conversion of NTemplate

Postby nurzico » Sep 28, 2015 09:00

Is it possible to extract template compatible of SDK version 3.1 using SDK version 4.3?
Or is there any way to convert template extracted from SDK version 3.1 to SDK version 4.3?

We are having template extracted from SDK version 3.1. And currently using SDK version 4.3. We want to use the previously extracted templates to current SDK.

Admin: merged to "Matching questions"
nurzico
 
Posts: 25
Joined: Sep 25, 2013 07:36

Re: Conversion of NTemplate

Postby Martynas » Sep 28, 2015 10:27

nurzico wrote:Is it possible to extract template compatible of SDK version 3.1 using SDK version 4.3?
Or is there any way to convert template extracted from SDK version 3.1 to SDK version 4.3?

We are having template extracted from SDK version 3.1. And currently using SDK version 4.3. We want to use the previously extracted templates to current SDK.

Hello,

MegaMatcher 3.1 and 4.3 version templates are compatible with each other, so there is no need of conversion.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

How to Fuse biometrics SDK 5 (NMatcher deprecated)

Postby Lalopez » Oct 13, 2015 19:39

Hi,

I am developing a .NET application that uses the SDK 5.

The template used consists of the Face and two fingerprints; everything is working fine, except that when I try to identify someone with those 3 biometrics, I would expect that these are fusioned and the result be unique.
Even if I raise the MatchingThreshold of the BiometricClient object, in the vast majority it is returning more than one match.
In previous versions there was the NMatcher with the NMFusionType parameter, now the NMatcher object and all of its properties are deprecated.

I think that the default is Detect by Face then fuse, how can I change this so the result is created with all of the biometrices fused?

Thanks

Admin: merged to "Matching questions"
Lalopez
 
Posts: 3
Joined: Jul 03, 2015 16:19

Re: How to Fuse biometrics SDK 5 (NMatcher deprecated)

Postby Martynas » Oct 15, 2015 07:45

Lalopez wrote:Hi,

I am developing a .NET application that uses the SDK 5.

The template used consists of the Face and two fingerprints; everything is working fine, except that when I try to identify someone with those 3 biometrics, I would expect that these are fusioned and the result be unique.
Even if I raise the MatchingThreshold of the BiometricClient object, in the vast majority it is returning more than one match.
In previous versions there was the NMatcher with the NMFusionType parameter, now the NMatcher object and all of its properties are deprecated.

I think that the default is Detect by Face then fuse, how can I change this so the result is created with all of the biometrices fused?

Thanks

Hello,

the default fusion mode is "FuseAllways".
If you are getting issues with matches, then provide details on the issue with used parameters and templates to support@neurotechnology.com.
Best regards,

Martynas
Software Consultant
Neurotechnology
Martynas
Neurotechnology
 
Posts: 2192
Joined: Sep 19, 2008 10:02
Location: Vilnius, Lithuania

Face matching threshold problem

Postby pablo111 » Nov 17, 2015 16:49

I am using Neurotechnology SDK 4.4 for facial verification. And I want to get the real score out instead of zero, when the score is below 48. How can I change the default threshold?

Here is part of my C++ code:
Code: Select all
NResult result = N_OK;
   HNMatcher matcher = NULL;
   HNMatchingDetails hMatchingDetails = NULL;
   result = NMCreate(&matcher);
   if (NFailed(result))
   {
      return result;
   }
   result = NMVerifyEx(matcher, pTemplateIn1, nTemplateSizeIn1, pTemplateIn2, nTemplateSizeIn2, &hMatchingDetails, &nScoreOut);


Admin: merged to "Matching questions"
pablo111
 
Posts: 8
Joined: Nov 16, 2015 18:04

Re: Face matching threshold problem

Postby vaidasz » Nov 19, 2015 15:49

pablo111 wrote:I am using Neurotechnology SDK 4.4 for facial verification. And I want to get the real score out instead of zero, when the score is below 48. How can I change the default threshold?

Here is part of my C++ code:
Code: Select all
NResult result = N_OK;
   HNMatcher matcher = NULL;
   HNMatchingDetails hMatchingDetails = NULL;
   result = NMCreate(&matcher);
   if (NFailed(result))
   {
      return result;
   }
   result = NMVerifyEx(matcher, pTemplateIn1, nTemplateSizeIn1, pTemplateIn2, nTemplateSizeIn2, &hMatchingDetails, &nScoreOut);

Hello,

To set matching threshold use:
Code: Select all
HNMatcher matcher;
NInt threshold = 0;
result =  NObjectSetParameterEx(matcher, NMP_MATCHING_THRESHOLD, N_TYPE_INT, &threshold, sizeof(threshold));
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

Re: Matching questions

Postby Ridham Shah » Nov 20, 2015 08:55

Give me The solution of Matching process.error occurred in matching process is One or more errors occurred.what is solution for that?
Ridham Shah
 
Posts: 2
Joined: Nov 05, 2015 08:01

Re: Face matching threshold problem

Postby pablo111 » Nov 20, 2015 17:29

vaidasz wrote:To set matching threshold use:
Code: Select all
HNMatcher matcher;
NInt threshold = 0;
result =  NObjectSetParameterEx(matcher, NMP_MATCHING_THRESHOLD, N_TYPE_INT, &threshold, sizeof(threshold));

Thank you, Vaidas. It works!
pablo111
 
Posts: 8
Joined: Nov 16, 2015 18:04

Re: Matching questions

Postby vaidasz » Nov 24, 2015 14:35

Ridham Shah wrote:Give me The solution of Matching process.error occurred in matching process is One or more errors occurred.what is solution for that?


Hello Ridham,

The exception message you specified is incomplete. Please provide full exception message.
Vaidas Z
Neurotechnology
vaidasz
Neurotechnology
 
Posts: 559
Joined: Apr 30, 2010 06:57
Location: Vilnius, Lithuania

Re: Face matching threshold problem

Postby pradeep » Nov 28, 2015 11:29

Dear Sir,

What is the difference between Face-Enrollment and Face Identification both are creating templates but I want to save the template i Sqlite Database So which tutorial should i use?

And In Face verification we are using both Images How to verify Image with Template (OR) I have to convert New Image with Template or Old template to Image please.

Because I want to compare Image with template stored in database.

Regards;
Pradeep Khaire
pradeep
 
Posts: 8
Joined: Nov 24, 2015 13:16

Re: Face matching threshold problem

Postby pradeep » Nov 28, 2015 12:32

Dear Sir,

How template get saved in SQLite database becasue biometricclient.setSQLiteDatabseConnection(String args0) will create only file at specified filename.
Is that this method also saves Template in database??
Because we are not passing any template.
I not able to understand.

Because when I run abis-sample code during enrolment template get saved in database, Is my observation right ?

please confirm ..

Regards;
Pradeep Khaire
pradeep
 
Posts: 8
Joined: Nov 24, 2015 13:16

PreviousNext

Return to Public

Who is online

Users browsing this forum: No registered users and 2 guests