Methods and systems for generating shape data for electronic designs

1. A method for generation of shape data for a set of electronic designs, the method comprising:

inputting a set of shape data, wherein the set of shape data represents a set of shapes for a device fabrication process; and

using a convolutional neural network on the set of shape data to determine a set of generated shape data, wherein the convolutional neural network comprises a generator trained with a pre-determined set of discriminators;

wherein the set of generated shape data comprises a scanning electron microscope (SEM) image;

wherein the pre-determined set of discriminators comprises: a first pre-determined discriminator trained with 1024×1024 pixel images, a second pre-determined discriminator trained with 512×512 pixel images; and a third pre-determined discriminator trained with 256×256 pixel images.

2. The method of claim 1 wherein the set of shape data is created using lithography simulation.

3. The method of claim 1 wherein each pre-determined discriminator in the pre-determined set of discriminators outputs a discriminator loss and a perceptual loss.

4. The method of claim 3 wherein the discriminator losses from each of the pre-determined discriminators are combined together to form a final discriminator loss.

5. The method of claim 4 wherein a portion of the discriminator loss of each of the predetermined discriminators and the perceptual loss are combined to produce a final generator loss.

6. The method of claim 1 wherein each discriminator comprises a plurality of classifier blocks.

7. The method of claim 6 wherein the plurality of classifier blocks comprises:

a first block comprising a convolution layer, a leaky Rectified Linear Unit (ReLU) activation layer and a padding layer;

a second block, a third block, a fourth block, a fifth block and a sixth block each comprising a convolution layer, a batch normalization layer, a leaky ReLU activation layer and a padding layer; and

a seventh block comprising a convolution layer, a padding layer and a sigmoid activation layer.

8. The method of claim 1 further comprising inputting an actual SEM image.

9. The method of claim 8 wherein the actual SEM image is used by the pre-determined set of discriminators to compare with the set of generated shape data.

10. The method of claim 1 further comprising:

using the generator to create the set of generated shape data, wherein the generator comprises an encoder and a decoder.

11. The method of claim 10 wherein the encoder further comprises a plurality of encoder blocks, wherein each encoder block after a first encoder block in the plurality of encoder blocks comprises a batch normalization layer, a convolutional layer and a leaky ReLU activation layer.

12. The method of claim 10 wherein the decoder further comprises a plurality of decoder blocks, wherein:

an initial decoder block in the plurality of decoder blocks comprises a transpose convolution layer and a ReLU activation layer; followed by

a first set of decoder blocks comprising a transpose convolution layer, a batch normalization layer and a ReLU activation layer; and

a second set of decoder blocks comprising a transpose convolution layer, a batch normalization layer, a dropout layer and a ReLU activation layer.

13. The method of claim 1 wherein the device fabrication process is a semiconductor fabrication process.

14. The method of claim 1 wherein the device fabrication process is a flat panel display fabrication process.

15. The method of claim 1 wherein the set of shape data further comprises a simulated mask image.

16. A method for generation of a SEM image for a set of electronic designs, the method comprising:

inputting a set of shape data, wherein the set of shape data represents a set of shapes for a device fabrication process;

inputting a set of parameters including a set of convolution layers for a Conditional Generative Adversarial Network (CGAN) comprising a generator and a set of discriminators;

generating a SEM image with the set of shape data, using the set of convolution layers of the CGAN;

calculating a generator loss comprising a perceptual loss combined with a portion of a discriminator loss; and

adjusting the set of parameters including the set of convolution layers;

wherein the set of discriminators comprises: a first discriminator receiving 1024×1024 pixel images, a second discriminator receiving 512×512 pixel images, and a third discriminator receiving 256×256 pixel images.

17. The method of claim 16 wherein the set of parameters comprises a kernel size of 4×4 with channels varying from 3, 64, 128, 256 and 512 for each convolution layer.

18. The method of claim 16 wherein the portion of the discriminator loss is combined from the set of discriminators.

19. The method of claim 16 wherein the generator further comprises a U-net.