Method of training semiconductor process image generator

Patent metadata

Field	Value
Publication number	US-12561958-B2
Application number	US-202217820911-A
Country	US
Kind code	B2
Filing date	Aug 19, 2022
Priority date	Dec 6, 2021
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

A method of training a semiconductor process image generator includes training the semiconductor process image generator with a plurality of mask images including a first group and a second group, training the semiconductor process image generator with the second group and a first transformed group obtained by applying a transformation to the first group, and training the semiconductor process image generator with the first group and a second transformed group obtained by applying a transformation to the second group.

First claim

What is claimed is:

1 . A method of training a semiconductor process image generator, the method comprising:
training the semiconductor process image generator with a plurality of mask images including a first group and a second group, during a first time period; training the semiconductor process image generator with the second group and a first transformed group obtained by applying a transformation to the first group, during a second time period that is different from the first time period; and training the semiconductor process image generator with the first group and a second transformed group obtained by applying a transformation to the second group, during a third time period that is different from the first time period and the second time period.

2 . The method of claim 1 , wherein the transformation includes one of a rotation transformation, a symmetry transformation, and a combination of the rotation transformation and the symmetry transformation.

3 . The method of claim 1 , wherein
the training of the semiconductor process image generator with the plurality of mask images is repeated n1 times, the training of the semiconductor process image generator with the first transformed group and the second group is repeated n2 times, and n2 is less than or equal to n1.

4 . The method of claim 3 , wherein
the training of the semiconductor process image generator with the second transformed group and the first group is repeated n3 times, and n3 is less than or equal to n1.

5 . The method of claim 1 , wherein
the training of the semiconductor process image generator with the plurality of mask images is repeated until a loss function of the semiconductor process image generator is less than or equal to a first value, the training of the semiconductor process image generator with the first transformed group and the second group is repeated until the loss function of the semiconductor process image generator is less than or equal to a second value, and the second value is less than or equal to the first value.

6 . The method of claim 5 , wherein
the training of the semiconductor process image generator with the second transformed group and the first group is repeated until the loss function of the semiconductor process image generator is less than or equal to a third value, and the third value is less than or equal to the first value.

7 . The method of claim 1 , wherein the semiconductor process image generator comprises a plurality of convolution layers.

8 . A method of training a semiconductor process image generator, the method comprising:
training the semiconductor process image generator with a first mask image and a second mast image, during a first time period; transforming the first mask image to obtain a third mask image; training the semiconductor process image generator with the second mask image and the third mask image, during a second time period following the first time period; transforming the second mask image to obtain a fourth mask image; and training the semiconductor process image generator with the first mask image and the fourth mask image, during a third time period following the second time period.

9 . The method of claim 8 , wherein the first mask image is transformed by a rotation transformation.

10 . The method of claim 8 , wherein the first mask image is transformed by a symmetry transformation.

11 . The method of claim 8 , wherein the first mask image is transformed by a 90 degree rotation transformation.

12 . The method of claim 8 , wherein the first mask image is transformed by a 180 degree rotation transformation.

13 . The method of claim 8 , further comprising repeating the training the semiconductor process image generator with the first mask image and the second mast image, during a fourth time period that is different from the first time period, the second time period and the third time period.

14 . A method of training a semiconductor process image generator, the method comprising:
training the semiconductor process image generator with a first group and a second group, during a first time period, the first group including a first mask image and a second mask image, the second group including a third mask image and a fourth mask image; transforming the first mask image to obtain a first transformed mask image, and transforming the second mask image to obtain a second transformed mask image; training the semiconductor process image generator with the third mask image, the fourth mask image, the first transformed mask image and the second transformed mask image, during a second time period that is different from the first time period; transforming the third mask image to obtain a third transformed mask image, and transforming the fourth mask image to obtain a fourth transformed mask image; and training the semiconductor process image generator with the first mask image, the second mask image, the third transformed mask image and the fourth transformed mask image, during a third time period that is different from the first time period and the second time period.

15 . The method of claim 14 , wherein the first mask image and the second mask image are transformed by a rotation transformation.

16 . The method of claim 14 , wherein the first mask image and the second mask image are transformed by a symmetry transformation.

17 . The method of claim 14 , wherein the first mask image and the second mask image are transformed by a 90 degree rotation transformation.

18 . The method of claim 14 , wherein the first mask image and the second mask image are transformed by a 180 degree rotation transformation.

19 . The method of claim 14 , further comprising repeating the training the semiconductor process image generator with the first group and the second group, during a fourth time period that is different from the first time period, the second time period and the third time period.

20 . The method of claim 14 , further comprising randomly selecting the first mask image and the second mask image among a plurality of mask images to form the first group.

Assignees

• Samsung Electronics Co Ltd

Inventors

• Yeo Sangchul
• Yang Jaewon
• Lee Hyeok
• Lee Sooryong

Classifications

• G06T3/60

  Rotation of whole images or parts thereof · CPC title

• G06V10/774 Primary

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06V10/776 Primary

  Validation; Performance evaluation · CPC title

• G06V10/242

  by image rotation, e.g. by 90 degrees · CPC title

• G03F7/705

  Modelling or simulating from physical phenomena up to complete wafer processes or whole workflow in wafer productions · CPC title