Computer system, dimension measuring method, and semiconductor device manufacturing system

A dimension measuring method of a computer system extracting coordinate information on base points for measuring a dimension of a desired portion of image data, measuring the dimension by using the coordinate information on the base points, wherein

• a preprocessing unit is provided which is configured to allow training by integrating, in a case where a plurality of pieces of training data having different number of measurement portions are included in a training data set, an image of the data in which the number of measurement portions is insufficient by setting an insufficient base point as an insufficient measurement portion and shielding a region assumed as the insufficient measurement portion,
• the preprocessing unit automatically designs a skeleton that includes at least two of the base points of the image data as key points, inputs image data to be measured into a pose estimation model trained to output coordinate information on each of the key points, and generates coordinate information on each of the key points of the input image data,
• the dimension is measured by using the coordinate information on the key points of the image data to be measured, and
• the pose estimation model is trained by using training data in which the image data is set as an input and the coordinate information on at least two of the base points is set as an output.

A semiconductor device manufacturing system provided with a platform in which an application for extracting coordinate information on base points for measuring a dimension of a desired portion of image data, and measuring the dimension by using the coordinate information on the base points is installed, wherein

• the application is configured to perform:
• a step of allowing training by integrating, in a case where a plurality of pieces of training data having different number of measurement portions are included in a training data set, an image of the data in which the number of measurement portions is insufficient by setting an insufficient base point as an insufficient measurement portion and shielding a region assumed as the insufficient measurement portion;
• a step of automatically designing a skeleton that includes at least two of the base points of the image data as key points, inputting image data to be measured into a pose estimation model trained to output coordinate information on each of the key points, and generating coordinate information on each of the key points of the input image data;
• and
• step of measuring the dimension by using the coordinate information on each of the key points of the image data to be measured, and
• the pose estimation model is trained by using training data in which the image data is set as an input and the coordinate information on at least two of the base points is set as an output.