Method for multiclass classification in open-set scenarios and uses thereof
The proposed method is used for classification in open-set scenarios, wherein often it is not possible to first obtain the training data for all possible classes that may arise during the testing stage. During the test phase, test samples belonging to one of the classes used in the training phase are classified based on a ratio between similarity scores, as known correct class and test samples belonging to any other class are to be rejected and classified as unknown.
1. A method for multi-class classification of open-set scenarios of at least one of image recognition and speech recognition, the method comprising:
dividing, by at least one processor, a set of training samples stored in at least one memory, including at least one of an image sample and an audio sample, into classes of interest;
training, by the at least one processor, a multiclass classifier using the training samples divided into classes of interest, to obtain intrinsic parameters of the training samples;
acquiring, by the at least one processor, a rejection parameter through parameter optimization of the obtained intrinsic parameters;
receiving, by the at least one processor, a test sample including at least one of an image sample and an audio sample;
determining, by the at least one processor, two classes of the classes of interest that are similar to the test sample as a first class and a second class;
determining, by the at least one processor, a first similarity score between the first class and the test sample, and a second similarity score between the second class and the test sample;
determining, by the at least one processor, a ratio as the first similarity score divided by the second similarity score;
classifying, by the at least one processor, the test sample as:
a known sample if the determined ratio is less than the acquired rejection parameter, and
an unknown sample if the determined ratio is greater than or equal to the acquired rejection parameter;
providing, by the at least one processor, information related to the classified test sample based on the classification of the test sample, wherein the information includes one of an identification of the known sample and a request for a user to provide an identification of the unknown sample; and
applying, by the at least one processor, the provided information including at least one of the identification of the known sample and the identification of the unknown sample to the multiclass classifier for the classification of open-set scenarios to recognize the at least one of the image sample and the audio sample.
2. The method according to claim 1 , wherein the dividing the set of training samples comprises dividing the set of training samples into a fitting set and a validation set.
3. The method according to claim 2 , wherein the parameter optimization comprises uses the set of training samples included in the fitting set to simulate an open-set scenario, and executes a traditional grid search procedure to optimize the rejection parameter using the set of training samples included in the validation set.
4. The method according to claim 1 , wherein the determining the first class and the second class comprises finding two classes of the classes of interest that are most similar to the test sample.
5. The method according to claim 1 , wherein the rejection parameter is greater than or equal to zero and less than or equal to one.
6. The method according to claim 1 , further comprising identifying information of an unknown sample.
7. The method according to claim 5 , wherein the rejection parameter defines a bounded region in a characteristics space in which the test sample is classified as one of the known sample and an unlimited region in which the test sample is classified as the unknown sample.
8. The method according to claim 1 , wherein the first similarity score and the second similarity score include at least one of a cost function of Optimum-Path Forest classification and a distance metric of k-Nearest Neighbors classification.
9. The method according to claim 1 , wherein the test sample includes at least one of biometric information, facial information, object information, scene information, voice information, character information, sensor information, image information, video information, text information, speech information, audio information, and pattern information.
10. The method according to claim 1 , wherein the test sample includes at least one of medical information.
11. A non-transitory computer-readable recording medium storing a program to implement a method for multi-class classification of open-set scenarios of at least one of image recognition and speech recognition, the method comprising:
dividing, by at least one processor, a set of training samples stored in at least one memory, including at least one of an image sample and an audio sample, into classes of interest;
training, by the at least one processor, a multiclass classifier using the training samples divided into classes of interest, to obtain intrinsic parameters of the training samples;
acquiring, by the at least one processor, a rejection parameter through parameter optimization of the obtained intrinsic parameters;
receiving, by the at least one processor, a test sample including at least one of an image sample and an audio sample;
determining, by the at least one processor, two classes of the classes of interest that are similar to the test sample as a first class and a second class;
determining, by the at least one processor, a first similarity score between the first class and the test sample, and a second similarity score between the second class and the test sample;
determining, by the at least one processor, a ratio as the first similarity score divided by the second similarity score;
classifying, by the at least one processor, the test sample as:
a known sample if the determined ratio is less than the acquired rejection parameter, and
an unknown sample if the determined ratio is greater than or equal to the acquired rejection parameter;
providing, by the at least one processor, information related to the classified test sample based on the classification of the test sample, wherein the information includes one of an identification of the known sample and a request for a user to provide an identification of the unknown sample; and
applying, by the at least one processor, the provided information including at least one of the identification of the known sample and the identification of the unknown sample to the multiclass classifier for the classification of open-set scenarios to recognize the at least one of the image sample and the audio sample.
12. An apparatus comprising:
at least one memory configured to store instructions; and
at least one processor configured to execute the stored instructions to implement a method for multi-class classification of open-set scenarios of at least one of image recognition and speech recognition, the method comprising:
dividing, by the at least one processor, a set of training samples stored in the at least one memory, including at least one of an image sample and an audio sample, into classes of interest;
training, by the at least one processor, a multiclass classifier using the training samples divided into classes of interest, to obtain intrinsic parameters of the training samples;
acquiring, by the at least one processor, a rejection parameter through parameter optimization of the obtained intrinsic parameters;
receiving, by the at least one processor, a test sample including at least one of an image sample and an audio sample;
determining, by the at least one processor, two classes of the classes of interest that are similar to the test sample as a first class and a second class;
determining, by the at least one processor, a first similarity score between the first class and the test sample, and a second similarity score between the second class and the test sample;
determining, by the at least one processor, a ratio as the first similarity score divided by the second similarity score;
classifying, by the at least one processor, the test sample as:
a known sample if the determined ratio is less than the acquired rejection parameter, and
an unknown sample if the determined ratio is greater than or equal to the acquired rejection parameter;
providing, by the at least one processor, information related to the classified test sample based on the classification of the test sample, wherein the information includes one of an identification of the known sample and a request for a user to provide an identification of the unknown sample; and
applying, by the at least one processor, the provided information including at least one of the identification of the known sample and the identification of the unknown sample to the multiclass classifier for the classification of open-set scenarios to recognize the at least one of the image sample and the audio sample.