Jeans with laser finishing patterns created by neural network

Patent metadata

Field	Value
Publication number	US-12344979-B2
Application number	US-202217651210-A
Country	US
Kind code	B2
Filing date	Feb 15, 2022
Priority date	Oct 31, 2017
Publication date	Jul 1, 2025
Grant date	Jul 1, 2025

Abstract

Software and lasers are used in finishing apparel to produce a desired wear pattern or other design. A technique includes using machine learning to create or extract a laser input file for wear pattern from an existing garment. Machine learning can be by a generative adversarial network, having generative and discriminative neural nets. The generative adversarial network is trained and then used to create a model. This model is used generate the laser input file from an image of the existing garment with the finishing pattern. With this laser input file, a laser can re-create the wear pattern from the existing garment onto a new garment.

First claim

The invention claimed is:

1. A method of manufacturing a pair of jeans comprising:
providing a target pair of jeans comprising fabric panels made from a material comprising a warp comprising dyed yarn, wherein the fabric panels are sewn together using thread;
using a laser with a laser input file as input to create on an outer surface of the target jeans a finishing pattern,
providing an image of an existing pair of jeans with finishing pattern;
providing a generative adversarial network comprising a generative neural net and a discriminative neural net;
forming a model from the generative adversarial network; and
using the model to generate the laser input file from the image of the existing garment with the finishing pattern, wherein the laser input file comprises digital data that is representative of the finishing pattern of the existing garment.

2. The method of claim 1 wherein before being exposed to the laser, a cross section of the warp of the target jeans comprises a generally round shape, and after being exposed to the laser, a depth of material has been removed, and a cross section of the warp comprises a region with a flattened shape relative to the previous generally round shape.

3. The method of claim 1 comprising:
providing a plurality of images of a plurality of sample garments with lasered finishing patterns resulting from a plurality of sample laser input files; and
inputting the sample laser input files and images of sample garments with lasered finishing patterns to the generative adversarial network, wherein the sample laser input files comprise real laser input files.

4. The method of claim 3 comprising:
using a generative neural net of the generative adversarial network to generate fake laser input files for the images of sample garments with lasered finishing patterns;
determining a generator loss based on the fake laser input files and real laser input files;
inputting the real laser input files to a real discriminator and a fake discriminator of a generative adversarial network;
inputting the fake laser input files to the fake discriminator of the generative adversarial network; and
determining a discriminator loss based on outputs of the real discriminator and fake discriminator.

5. A method of manufacturing a garment comprising:
providing a plurality of images of a plurality of sample garments with lasered finishing patterns resulting from a plurality of sample laser input files;
inputting the plurality of images to a generative adversarial network to iteratively train a final model;
inputting to the final model an image of an existing garment with a desired finishing pattern;
obtaining from the final model an output of a generated laser input file, which can be used to create the designed finishing pattern;
providing a target garment comprising fabric panels made from a first material comprising a warp comprising dyed cotton yarn, wherein the fabric panels are sewn together using thread;
inputting the generated laser input file to a laser; and
using the laser input file to control the laser to form the desired finishing pattern on a surface of the target garment.

6. The method of claim 5 wherein before the desired finishing pattern is formed, a cross section of the warp of the target garment comprises a generally round shape, and
after being exposed to the laser, a depth of material has been removed, and a cross section of the warp comprises a region with a flattened shape relative to the previous generally round shape.

7. The method of claim 5 wherein a preexisting laser input file for the desired finishing pattern does not exist.

8. The method of claim 5 wherein each laser input file comprises digital data that is representative of a finishing pattern on a sample garment.

9. The method of claim 5 comprising:
inputting the sample laser input files and images of sample garments with lasered finishing patterns to the generative adversarial network, wherein the sample laser input files comprise real laser input files; and
using a generative neural net of the generative adversarial network to generate fake laser input files for the images of sample garments with lasered finishing patterns.

10. The method of claim 9 comprising:
determining a generator loss based on the fake laser input files and real laser input files.

11. The method of claim 10 comprising:
inputting the real laser input files to a real discriminator and a fake discriminator of a generative adversarial network;
inputting the fake laser input files to the fake discriminator of the generative adversarial network; and
determining a discriminator loss based on outputs of the real discriminator and fake discriminator.

12. The method of claim 11 wherein a model is iteratively trained to obtain a final model based on outputs of the generator loss and discriminator loss.

13. The method of claim 9 comprising:
inputting the real laser input files to a real discriminator and a fake discriminator of a generative adversarial network;
inputting the fake laser input files to the fake discriminator of the generative adversarial network; and
determining a discriminator loss based on outputs of the real discriminator and fake discriminator.

14. The method of claim 5 wherein the generated laser input file is for a finishing pattern on an existing garment, and
the existing garment made from a second material, wherein the second material comprises a different fabric characteristic from the first material, and the existing garment was created before the target garment was created.

15. The method of claim 5 wherein the warp is ring dyed using an indigo dye,
based on the generated laser input file, selected amounts of material are removed by the laser from the surface of the first material at different pixel locations of the target garment, and
for lighter pixel locations of the finishing pattern, a greater depth of the dyed cotton warp yarn is removed, revealing a greater width of an inner core of the dyed yarn, while for darker pixel locations of the finishing pattern, a lesser depth of the dyed cotton warp yarn is removed, revealing a lesser width of an inner core of the dyed yarn.

16. The method of claim 5 wherein the finishing pattern created can extend across portions of the target garment where two or more fabric panels are joined together by thread by exposing these portions to the laser.

17. The method of claim 5 wherein the target garment comprises a pair of jeans.

18. A method of manufacturing a garment comprising:
providing a target garment comprising fabric panels made from a material comprising a warp comprising dyed yarn, wherein the fabric panels are sewn together using thread;
using a laser with a laser input file as input to create on an outer surface of the target garment a finishing pattern,
wherein before being exposed to the laser, a cross section of the warp of the target garment comprises a generally round shape, and after being exposed to the laser, a depth of material has been removed, a cross section of the warp comprises a region with a flattened shape relative to the previous generally round shape;
providing an image of an existing garment with finishing pattern;
providing a generative adversarial network comprising a generative neural net and a discriminative neural net;
forming a model from the generative adversarial network; and
using the model to generate the laser input file from the image of the existing garment with the finishing pattern, wherein the laser input file comprises dig

Assignees

• Strauss Levi & Co

Inventors

• Mahanty Debdulal
• Schultz Christopher
• Schultz Jennifer
• Bell Benjamin

Classifications

• G06N3/0475

  Generative networks · CPC title

• G06N3/094

  Adversarial learning · CPC title

• D06B11/0096 Primary

  to get a faded look · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/088

  Non-supervised learning, e.g. competitive learning · CPC title