Data recognition model construction apparatus and method for constructing data recognition model thereof, and data recognition apparatus and method for recognizing data thereof

1. A data recognition model construction apparatus, comprising:

a video inputter configured to receive a video;

an image composition unit configured to obtain at least one common object included in each of a plurality of images that form at least a portion of the video, and obtain a composition image by overlaying a plurality of common areas corresponding to the at least one common object with each other, each of the plurality of common areas being included in a corresponding image among the plurality of images and including the at least one common object;

a learning data inputter configured to receive the obtained composition image;

a model learning unit configured to obtain learning data comprising the composition image and a classification result corresponding to the composition image, the model learning unit being further configured to make a data recognition model learn using the obtained composition image and the classification result; and

a model storage configured to store the learnt data recognition model,

wherein the image composition unit is configured to determine the plurality of common areas, each of the common areas having a size of a motion being greater than or equal to a predetermined vector, from the plurality of images.

2. The apparatus as claimed in claim 1 , wherein the image composition unit is configured to extract a common area from each of the plurality of images and obtain the composition image by overlaying the plurality of extracted common areas.

3. The apparatus as claimed in claim 1 , wherein each of the plurality of common areas is configured to include an identical or similar common object from each of the plurality of images.

4. The apparatus as claimed in claim 1 , wherein the data recognition model construction apparatus further comprises a model evaluation unit configured to evaluate the learnt data recognition model using the video or the composition image.

5. The apparatus as claimed in claim 1 ,

wherein the data recognition model is a neural network model or a deep learning model which performs computing based on a connection relation among a plurality of network nodes and a weighted value of each of the plurality of network nodes.

6. A data recognition apparatus, comprising:

a video inputter configured to receive a video;

an image composition unit configured to obtain at least one common object included in each of a plurality of images that form at least a portion of the video, and obtain a composition image by overlaying a plurality of common areas corresponding to the at least one common object with each other, each of the plurality of common areas being included in a corresponding image among the plurality of images and including the at least one common object; and

a recognition result provider configured to provide a recognition result of the video by applying the obtained composite image to a data recognition model,

wherein the image composition unit is configured to determine the plurality of common areas, each of the common areas having a size of a motion being greater than or equal to a predetermined vector from the plurality of images, and

wherein the data recognition model is learned based on learning data comprising the composition image and a classification result corresponding to the composition image.

7. The apparatus as claimed in claim 6 , wherein the image composition unit is configured to extract a common area from each of the plurality of images and obtain the composition image by overlaying the plurality of extracted common areas.

8. The apparatus as claimed in claim 6 , wherein each of the plurality of common areas is configured to include an identical or similar common object from each of the plurality of images.

9. The apparatus as claimed in claim 6 , wherein the recognition result of the video comprises state information of an object included in the video or information of a surrounding state of the object.

10. The apparatus as claimed in claim 6 , wherein the data recognition model is a neural network model or a deep learning model which performs computing based on a connection relation among a plurality of network nodes and a weighted value of each of the plurality of network nodes.

11. The apparatus as claimed in claim 6 , wherein the image composition unit obtains the composition image based on a center of gravity of the common area.

12. The apparatus as claimed in claim 6 , wherein the image composition unit obtains a composition image by overlaying the common area and a background area around the common area included in each of the plurality of images.

13. The apparatus as claimed in claim 6 , wherein the image composition unit is configured to compensate a motion of a camera which photographs the video from each of the plurality of images, and obtain a composition image based on a common area included in each of the plurality of motion compensated images.

14. The apparatus as claimed in claim 6 , further comprising a model update unit which identifies whether the data recognition model is updated and updates the data recognition model according to an identification result.

15. The apparatus as claimed in claim 6 , further comprising a service provider which provides a service based on a result of recognition of the video.

16. A method for constructing a data recognition model by a data recognition model construction apparatus, the method comprising:

receiving a video;

obtaining at least one common object included in each of a plurality of images that form at least a portion of the video;

obtaining a composition image by overlaying a plurality of common areas corresponding to the at least one common object with each other, each of the plurality of common areas being included in a corresponding image among the plurality of images and including the at least one common object;

obtaining learning data comprising the composition image and a classification result corresponding to the composition image;

learning a data recognition model using the obtained composition image and the classification result; and

storing the learnt data recognition model,

wherein the obtaining the composition image comprises determining the plurality of common areas, each of the common areas having a size of a motion being greater than or equal to a predetermined vector from the plurality of images.

17. The method as claimed in claim 16 , wherein the obtaining a composition image comprises:

extracting a common area from each of the plurality of images; and

obtaining a composition image by overlaying the plurality of extracted common areas.

18. A method for recognizing data by a data recognition apparatus, the method comprising:

receiving a video;

obtaining at least one common object included in each of a plurality of images that form at least a portion of the video;

obtaining a composition image by overlaying a plurality of common areas corresponding to the at least one common object with each other, each of the plurality of common areas being included in a corresponding image among the plurality of images and including the at least one common object; and

applying the obtained composition image to a data recognition model and providing a recognition result of the video,

wherein the obtaining the composition image comprises determining the plurality of common areas, each of the common areas having a size of a motion being greater than or equal to a predetermined vector from the plurality of images, and

wherein the data recognition model is learned based on learning data comprising the composition image and a classification result corresponding to the composition image.

19. The method as claimed in claim 18 , wherein the obtaining a composition image comprises:

extracting a common area from each of the plurality of images; and

obtaining the composition image by overlaying the plurality of extracted common areas.

20. The method as claimed in claim 18 , wherein each of the plurality of common areas is configured to include an identical or similar common object from each of the plurality of images.