Segmentation of target areas in images

1. A computer-implemented method comprising:

receiving an input image that includes a target area;

generating, using two different extracting models, two different segments for extracting the same target area from the input image;

comparing the two different segments to determine a combined segment of at least the target area.

2. The computer-implemented method of claim 1 , further comprising:

extracting the target area from the input image using the combined segment; and

classifying the input image based on the target area extracted from the input image.

3. The computer-implemented method of claim 1 , wherein the two different extracting models include a first and second extracting model and the two different segments include a first and second segment, further comprising generating, using a third extracting model, a third segment for extracting the target area from the input image, wherein:

the comparing the two different segments includes comparing respective pixels of the first segment to respective pixels of the second segment to respective pixels of the third segment to determine a value of each pixel in the combined segment based on majority rule.

4. The computer-implemented method of claim 1 , wherein the comparing the two different segments includes determining a value of each pixel in the combined segment based on an average of values of corresponding pixels in the two different segments.

5. The computer-implemented method of claim 1 , wherein the comparing includes generating the combined segment using a combining model, and the combining model inputs the two different segments to generate the combined segment.

6. The computer-implemented method of claim 5 , further comprising training the combining model based on at least one training dataset, wherein the training dataset includes a plurality of training segments and a training combined segment.

7. The computer-implemented method of claim 6 , wherein the generating the plurality of segments includes generating, using the two different extracting models, the plurality of training segments for extracting a target area in a training input image.

8. The computer-implemented method of claim 5 , wherein the combining model includes a neural network.

9. The computer-implemented method of claim 1 , further comprising training the two different extracting models based on training datasets, wherein each training dataset includes a training input image and a training segment.

10. The computer-implemented method of claim 9 , wherein the training the two different extracting models includes training each extracting model using a K-fold cross-validation method.

11. The computer-implemented method of claim 10 , wherein the training the two different extracting models includes:

dividing the training datasets into a plurality of subsets, and

assigning, for each extracting model, a different combination of subsets from the plurality of subsets for training and validating each extracting model.

12. The computer-implemented method of claim 9 , wherein the training the two different extracting models includes:

training a first extracting model of the two different extracting models; and

training a second extracting model of the two different extracting models, wherein the second extracting model is further trained from the first extracting model.

13. A computer program product including one or more computer readable storage mediums collectively storing program instructions that, when executed by a processor or programmable circuitry, cause the processor or programmable circuitry to perform operations comprising:

generating, using a first extracting model, a first segment for extracting a target area from an input image;

generating, using a second extracting model that is different than the first extracting model, a second segment for extracting the target area from the input image; and

comparing the first segment to the second segment to determine a combined segment of at least the target area.

14. The computer program product of claim 13 , further comprising instructions that, when executed by the processor or programmable circuitry, cause the processor or programmable circuitry to perform operations comprising generating, using a third extracting model, a third segment for extracting the target area from the input image, wherein:

the comparing the first segment to the second segment includes comparing respective pixels of the first segment to respective pixels of the second segment to respective pixels of the third segment to determine a value of each pixel in the combined segment based on majority rule.

15. The computer program product of claim 13 , wherein the comparing includes generating the combined segment using a combining model, and the combining model inputs the first segment and the second segment to generate the combined segment.

16. The computer program product of claim 15 , further comprising instructions that, when executed by the processor or programmable circuitry, cause the processor or programmable circuitry to perform operations comprising training the combining model based on at least one training dataset, wherein each training dataset includes a plurality of training segments and a training combined segment.

17. The computer program product of claim 13 , wherein the combining model includes a neural network.

18. The computer program product of claim 13 , further comprising instructions that, when executed by the processor or programmable circuitry, cause the processor or programmable circuitry to perform operations comprising training the first and second extracting models based on training datasets, wherein each training dataset includes a training input image and a training segment.

19. The computer program product of claim 18 , wherein the training the first and second extracting models includes training each extracting model using a K-fold cross-validation method.

20. An apparatus comprising:

a processor or a programmable circuitry; and

one or more computer readable mediums collectively including instructions that, when executed by the processor or the programmable circuitry, cause the processor or the programmable circuitry to:

generate, using a first extracting model, a first segment for extracting a target area from an input image;

generate, using a second extracting model that is different than the first extracting model, a second segment for extracting the target area from the input image; and

compare the first segment to the second segment to determine a combined segment of at least the target area.

21. The apparatus of claim 20 , further comprising instructions that, when executed by the processor or programmable circuitry, cause the processor or programmable circuitry to generate, using a third extracting model, a third segment for extracting the target area from the input image, wherein:

the comparing the first segment to the second segment includes comparing respective pixels of the first segment to respective pixels of the second segment to respective pixels of the third segment to determine a value of each pixel in the combined segment based on majority rule.

22. The apparatus of claim 20 , wherein the comparing includes generating the combined segment using a combining model, and the combining model inputs the first segment and the second segment to generate the combined segment.

23. The apparatus of claim 22 , further comprising instructions that, when executed by the processor or programmable circuitry, cause the processor or programmable circuitry to train the combining model based on at least one training dataset, wherein each training dataset includes a plurality of training segments and a training combined segment.

24. The apparatus of claim 20 , further comprising instructions that, when executed by the processor or programmable circuitry, cause the processor or programmable circuitry to train the first and second extracting models based on training datasets, wherein each training dataset includes a training input image and a training segment.

25. The apparatus of claim 24 , wherein the training the first and second extracting models includes training each extracting model using a K-fold cross-validation method.