Electronic apparatus and method for controlling thereof

1. A method for controlling an electronic apparatus, the method comprising:

obtaining, from an external server, a neural network model trained to detect an object corresponding to at least one class;

obtaining a user command for detecting a first object corresponding to a first class, the user command including at least one keyword associated with the first class; and

based on the first object not corresponding to the at least one class, obtaining, by the electronic apparatus, a new neural network model based on the neural network model obtained from the external server and information of the first object that is personalized based on a user of the electronic apparatus, wherein the new neural network model is stored on the electronic apparatus and is obtained by locally extending the neural network model on the electronic apparatus, to detect the first object corresponding to the first class in addition to detecting the object corresponding to the at least one class.

2. The method as claimed in claim 1 , wherein the trained neural network model comprises:

a feature extraction module configured to extract a feature value of an object; and

a classification value obtaining module configured to obtain a classification value of the object based on the extracted feature value.

3. The method as claimed in claim 2 , wherein:

the obtaining the user command comprises obtaining an image of the first object; and

the obtaining the new neural network model comprises:

obtaining a first feature value of the first object by inputting the obtained image of the first object to the feature extraction module, and

obtaining a new classification value obtaining module based on the obtained first feature value and the classification value obtaining module.

4. The method as claimed in claim 3 , wherein:

the classification value obtaining module comprises a weight vector comprising a plurality of column vectors; and

the obtaining the new classification value obtaining module comprises:

generating a first column vector based on an average value of the first feature value, and

obtaining the new classification value obtaining module by adding the first column vector as a new column vector of the weight vector.

5. The method as claimed in claim 4 , further comprising regularizing the obtained new classification value obtaining module based on a predefined regularizing function.

6. The method as claimed in claim 1 , wherein the trained neural network model is trained based on a loss function comprising a predefined regularizing function to prevent overfitting.

7. The method as claimed in claim 1 , further comprising:

based on the user command being obtained, determining whether the first object corresponds to the at least one class,

wherein the determining comprises:

obtaining an image of the first object,

obtaining a first feature value of the first object by inputting the image of the first object to the neural network model, and

determining whether the first object corresponds to the at least one class by comparing the first feature value with a weight vector of the neural network model.

8. An electronic apparatus comprising:

a memory comprising at least one instruction; and

a processor configured to execute the at least one instruction to:

obtain, from an external server, a neural network model trained to detect an object corresponding to at least one class,

obtain a user command for detecting a first object corresponding to a first class, the user command including at least one keyword associated with the first class, and

based on the first object not corresponding to the at least one class, obtain a new neural network model based on the neural network model obtained from the external server and information of the first object that is personalized based on a user of the electronic apparatus, wherein the new neural network model is stored on the electronic apparatus and is obtained by locally extending the neural network model on the electronic apparatus to detect the first object corresponding to the first class in addition to detecting the object corresponding to the at least one class.

9. The electronic apparatus as claimed in claim 8 , wherein the trained neural network model comprises:

a feature extraction module configured to extract a feature value of an object; and

a classification value obtaining module configured to obtain a classification value of the object based on the extracted feature value.

10. The electronic apparatus as claimed in claim 9 , wherein the processor is further configured to execute the at least one instruction to:

obtain an image of the first object;

obtain a first feature value of the first object by inputting the obtained image of the first object to the feature extraction module; and

obtain a new classification value obtaining module based on the obtained first feature value and the classification value obtaining module.

11. The electronic apparatus as claimed in claim 10 , wherein:

the classification value obtaining module comprises a weight vector comprising a plurality of column vectors; and

the processor is further configured to execute the at least one instruction to:

generate a first column vector based on an average value of the first feature value, and

obtain the new classification value obtaining module by adding the first column vector as a new column vector of the weight vector.

12. The electronic apparatus as claimed in claim 11 , wherein the processor is further configured to execute the at least one instruction to regularize the obtained new classification value obtaining module based on a predefined regularizing function.

13. The electronic apparatus as claimed in claim 8 , wherein the trained neural network model is trained based on a loss function comprising a predefined regularizing function to prevent overfitting.

14. The electronic apparatus as claimed in claim 8 , wherein the processor is further configured to execute the at least one instruction to:

obtain an image of the first object;

obtain a first feature value of the first object by inputting the image of the first object to the neural network model; and

determine whether the first object corresponds to the at least one class by comparing the first feature value with a weight vector of the neural network model.

15. A non-transitory computer-readable recording medium having recorded thereon instructions executable by at least one processor to perform a method for controlling an electronic apparatus, the method comprising:

obtaining, from an external server, a neural network model trained to detect an object corresponding to at least one class;

obtaining a user command for detecting a first object corresponding to a first class, the user command including at least one keyword associated with the first class; and

based on the first object not corresponding to the at least one class, obtaining, by the electronic apparatus, a new neural network model based on the neural network model obtained from the external server and information of the first object that is personalized based on a user of the electronic apparatus, wherein the new neural network model is stored on the electronic apparatus and is obtained by locally extending the neural network model on the electronic apparatus to detect the first object corresponding to the first class in addition to detecting the object corresponding to the at least one class.

16. The non-transitory computer-readable recording medium as claimed in claim 15 , wherein the trained neural network model comprises:

a feature extraction module configured to extract a feature value of an object; and

a classification value obtaining module configured to obtain a classification value of the object based on the extracted feature value.

17. The non-transitory computer-readable recording medium as claimed in claim 16 , wherein:

the obtaining the user command comprises obtaining an image of the first object; and

the obtaining the new neural network model comprises:

obtaining a first feature value of the first object by inputting the obtained image of the first object to the feature extraction module, and

obtaining a new classification value obtaining module based on the obtained first feature value and the classification value obtaining module.

18. The non-transitory computer-readable recording medium as claimed in claim 17 , wherein:

the classification value obtaining module comprises a weight vector comprising a plurality of column vectors; and

the obtaining the new classification value obtaining module comprises:

generating a first column vector based on an average value of the first feature value, and

obtaining the new classification value obtaining module by adding the first column vector as a new column vector of the weight vector.

19. The non-transitory computer-readable recording medium as claimed in claim 15 , wherein the trained neural network model is trained based on a loss function comprising a predefined regularizing function to prevent overfitting.

20. The non-transitory computer-readable recording medium as claimed in claim 15 , wherein the method further comprises:

based on the user command being obtained, determining whether the first object corresponds to the at least one class,

wherein the determining comprises:

obtaining an image of the first object,

obtaining a first feature value of the first object by inputting the image of the first object to the neural network model, and

determining whether the first object corresponds to the at least one class by comparing the first feature value with a weight vector of the neural network model.