Method for the classification of a biometric trait represented by an input image

1. A method for classifying a biometric trait represented by an input image, the method being characterized in that it comprises the implementation, by data processing means of a client, of the steps of:

(a) by means of a convolutional neural network (CNN), performing multi-label classification of the input image to determine for each general pattern of a predefined set of non-exclusive possible general patterns of biometric traits, a Boolean representative as to whether said biometric trait presents or not said general pattern.

2. The method according to claim 1 , wherein said CNN has at least one residual connection.

3. The method according to claim 2 , wherein said CNN comprises at least one dense block having all possible residual connections.

4. The method according to claim 3 , wherein said CNN comprises a plurality of said dense blocks, with a size of feature maps of the dense blocks, decreasing from the input to the output of the CNN and/or a number of layers per dense block increasing from the input to the output of the CNN.

5. The method according to claim 4 , wherein the rate of decrease in the size of the feature maps of the dense blocks increases from the input to the output of the CNN.

6. The method according to claim 1 , comprising a prior training step (a 0 ), by means of data processing means of a server, from an image database of pre-classified biometric traits, of parameters of said CNN.

7. The method according to claim 6 wherein said CNN has at least one residual connection and said CNN comprises at least one dense block having all possible residual connections, wherein the step (a 0 ) comprises the random deactivation of layers of said dense block.

8. The method defined by claim 7 wherein the random deactivation of layers of said dense block has a probability of about 10%.

9. The method according to claim 1 , comprising a step (b) for the processing of said input image in such a way as to extract the features of interest of the biometric trait represented by said input image.

10. The method according to claim 9 , wherein said biometric traits are fingerprints, the features of interest to be extracted from the at least one fingerprint represented by said input image comprising the position and/or orientation of minutia.

11. The method according to claim 9 , wherein the at least one biometric trait represented by the input image is that of an individual, the method further comprising a step (c) of identifying or authenticating said individual by comparing the features of interest extracted from the biometric trait represented by said input image, with the features of reference biometric traits having the general pattern or patterns determined in step (a) as presented by the biometric trait represented by the input image.

12. The method according to claim 1 , wherein said biometric traits are fingerprints, and said predefined set of possible general patterns comprises the following general patterns: left loop, right loop, arch and spiral.

13. A non-transitory computer-readable program product comprising code instructions for the execution of a method according to claim 1 for the classification of a biometric trait represented by an input image, when said program is executed on a computer.

14. A non-transitory storage means readable by computer equipment whereupon a computer program product comprises code instructions for the execution of a method according to claim 1 for the classification of a biometric trait represented by an input image.