Method and apparatus with semiconductor image processing

1 . A processor-implemented method, the method comprising:

identifying input semantic components of a semiconductor pattern of an original input image and output semantic components of a semiconductor pattern of an original output image respectively from the original input image corresponding to an application target of a process for manufacturing a semiconductor and the original output image corresponding to an application result of the process;

determining component pairs based on a matching relationship between the input semantic components and the output semantic components;

generating an augmented input image and an augmented output image by transforming a transformation target being a part of the input semantic components from the original input image and the original output image;

generating a result image based on:

a first execution result of a neural model for estimating pattern transformation according to the process generated by inputting the original input image to the neural model; and

a second execution result of the neural model generated by inputting the augmented input image to the neural model; and

training the neural model according to the first execution result and the second execution result.

2 . The method of claim 1 , wherein the generating of the augmented input image comprises generating the augmented input image by removing the transformation target.

3 . The method of claim 1 , wherein the transforming of the transformation target comprises any one or any combination of any two or more of removing the transformation target, scaling the transformation target, shifting the transformation target, and rotating the transformation target.

4 . The method of claim 1 , wherein a portion not corresponding to the transformation target of the original input image is maintained in the augmented input image.

5 . The method of claim 1 , wherein the identifying of the input semantic components comprises identifying, as an input semantic component of the input semantic components, a group of pixels comprising pixel values that are not zero in the original input image and that are connected to one another.

6 . The method of claim 1 , further comprising:

identifying the output semantic components of the semiconductor pattern of the original output image from the original output image corresponding to the application result of the process; and

generating the augmented output image by applying transformation corresponding to transformation of the transformation target of the original input image to the output semantic components of the original output image,

wherein the training of the neural model comprises training the neural model according to a difference between the augmented output image and the result image.

7 . The method of claim 1 , further comprising:

executing the neural model based on the original input image; and

estimating the pattern transformation according to the process by combining the first execution result of the neural model based on the original input image and the second execution result of the neural model based on the augmented input image.

8 . The method of claim 1 , wherein the process comprises either one or both of a development process and an etching process.

9 . A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors, configure the one or more processors to perform the method of claim 1 .

10 . A processor-implemented method, the method comprising:

identifying input semantic components of a semiconductor pattern of an original input image and output components of a semiconductor pattern of an original output image respectively from the original input image corresponding to an application target of a process for manufacturing a semiconductor and the original output image corresponding to an application result of the process;

determining component pairs based on a matching relationship between the input semantic components and the output components;

generating an augmented input image and an augmented output image by removing a transformation target being a part of the component pairs from the original input image and the original output image; and

training a neural model for predicting pattern transformation according to the process based on the augmented input image and the augmented output image.

11 . An apparatus, the apparatus comprising:

one or more processors; and

a memory comprising a non-transitory computer-readable storage medium storing instructions that, when executed by the one or more processors, configure the one or more processors to:

identify input semantic components of a semiconductor pattern of an original input image and output semantic components of a semiconductor pattern of an original output image respectively from the original input image corresponding to an application target of a process for manufacturing a semiconductor and the original output image corresponding to an application result of the process;

determine component pairs based on a matching relationship between the input semantic components and the output semantic components;

generate an augmented input image and an augmented output image by transforming a transformation target being a part of the input semantic components from the original input image and the original output image;

generate a result image based on:

a first execution result of a neural model for estimating pattern transformation according to the process generated by inputting the original input image to the neural model; and

a second execution result of the neural model generated by inputting the augmented input image to the neural model; and

train the neural model according to the first execution result and the second execution result.

12 . The apparatus of claim 11 , wherein, for the generating of the augmented input image, the instructions, when executed by the one or more processors, configure the one or more processors to generate the augmented input image by removing the transformation target.

13 . The apparatus of claim 11 , wherein, for the transforming of the transformation target, the instructions, when executed by the one or more processors, configure the one or more processors to perform any one or any combination of any two or more of removing the transformation target, scaling the transformation target, shifting the transformation target, and rotating the transformation target.

14 . The apparatus of claim 11 , wherein a portion not corresponding to the transformation target of the original input image is maintained in the augmented input image.

15 . The apparatus of claim 11 , wherein, for the identifying of the input semantic components, the instructions, when executed by the one or more processors, configure the one or more processors to identify, as one input semantic component of the input semantic components, a group of pixels comprising pixel values that are not zero in the original input image and that are connected to one another.

16 . The apparatus of claim 11 , wherein the instructions, when executed by the one or more processors, configure the one or more processors to:

identify output semantic components of the semiconductor pattern of the original output image from the original output image corresponding to the application result of the process;

generate the augmented output image by applying transformation corresponding to transformation of the transformation target of the original input image to the output semantic components of the original output image; and

train the neural model according to a difference between the augmented output image and the result image.

17 . The apparatus of claim 11 , wherein the instructions, when executed by the one or more processors, configure the one or more processors to:

execute the neural model based on the original input image; and

estimate the pattern transformation according to the process by combining the first execution result of the neural model based on the original input image and the second execution result of the neural model based on the augmented input image.

18 . The apparatus of claim 11 , wherein the process comprises either one or both of a development process and an etching process.