Generation of synthetic 3-dimensional object images for recognition systems

Patent metadata

Field	Value
Publication number	US-11574453-B2
Application number	US-202017012881-A
Country	US
Kind code	B2
Filing date	Sep 4, 2020
Priority date	Dec 15, 2015
Publication date	Feb 7, 2023
Grant date	Feb 7, 2023

Abstract

Techniques are provided for generation of synthetic 3-dimensional object image variations for training of recognition systems. An example system may include an image synthesizing circuit configured to synthesize a 3D image of the object (including color and depth image pairs) based on a 3D model. The system may also include a background scene generator circuit configured to generate a background for each of the rendered image variations. The system may further include an image pose adjustment circuit configured to adjust the orientation and translation of the object for each of the variations. The system may further include an illumination and visual effect adjustment circuit configured to adjust illumination of the object and the background for each of the variations, and to further adjust visual effects of the object and the background for each of the variations based on application of simulated camera parameters.

First claim

What is claimed is:

1. A processor-implemented method for generating 3-Dimensional (3D) object image variations, the method comprising:
rendering a plurality of color and depth (RGB-D) image pair variations of an object based on a synthesized RGB-D image pair of the object, ones of the RGB-D image pair variations associated with respective adjusted visual effects of a generated background scene, the visual effects of the generated background scene adjusted based on application of simulated camera parameters;
generating an object recognition classifier based on a first subset of the rendered RGB-D image pair variations; and
testing the object recognition classifier based on a second subset of the rendered RGB-D image pair variations.

2. The method of claim 1 , wherein the synthesized RGB-D image pair of the object is based on a 3D model of the object.

3. The method of claim 2 , wherein the 3D model of the object is generated by one or more of a computer aided design (CAD) tool, a 3D scanning tool configured to scan a physical sample of the object, or a 3D sculpting tool.

4. The method of claim 1 , wherein the simulated camera parameters are first simulated camera parameters, and the ones of the RGB-D image pair variations are associated with respective adjusted visual effects of the object, the visual effects of the object adjusted based on application of second simulated camera parameters.

5. The method of claim 1 , wherein the generated background scene includes a 2-Dimensional (2D) background scene randomly selected from a database of background scenes.

6. A system to render 3-Dimensional (3D) object image variations, the system comprising:
an image variation rendering circuit to render a plurality of color and depth (RGB-D) image pair variations of an object based on a synthesized RGB-D image pair of the object, ones of the RGB-D image pair variations associated with respective adjusted visual effects of a generated background scene, the visual effects of the generated background scene adjusted based on application of simulated camera parameters;
a machine learning system to generate an object recognition classifier based on a first subset of the rendered RGB-D image pair variations; and
a classifier testing system to test the object recognition classifier based on a second subset of the rendered RGB-D image pair variations.

7. The system of claim 6 , further including an image synthesizing circuit to synthesize the synthesized RGB-D image pair of the object based on a 3D model of the object, the 3D model of the object generated by one or more of a computer aided design (CAD) tool, a 3D scanning tool configured to scan a physical sample of the object, or a 3D sculpting tool.

8. The system of claim 6 , wherein the simulated camera parameters are first simulated camera parameters, and the ones of the RGB-D image pair variations are associated with respective adjusted visual effects of the object, the visual effects of the object adjusted based on application of second simulated camera parameters.

9. The system of claim 6 , further including a background scene generator circuit to randomly select the background scene from a database of 2-Dimensional (2D) background scenes.

10. At least one non-transitory computer readable storage medium comprising instructions that, when executed, cause one or more processors to at least:
render a plurality of color and depth (RGB-D) image pair variations of an object based on a synthesized RGB-D image pair of the object, ones of the RGB-D image pair variations associated with respective adjusted visual effects of a generated background scene, the visual effects of the generated background scene adjusted based on application of simulated camera parameters;
generate an object recognition classifier based on a first subset of the rendered RGB-D image pair variations; and
test the object recognition classifier based on a second subset of the rendered RGB-D image pair variations.

11. The computer readable storage medium of claim 10 , wherein the synthesized RGB-D image pair of the object is based on a 3-Dimensional (3D) model of the object.

12. The computer readable storage medium of claim 11 , wherein the 3D model of the object is generated by one or more of a computer aided design (CAD) tool, a 3D scanning tool configured to scan a physical sample of the object, or a 3D sculpting tool.

13. The computer readable storage medium of claim 10 , wherein the simulated camera parameters are first simulated camera parameters, and the ones of the RGB-D image pair variations are associated with respective adjusted visual effects of the object, the visual effects of the object adjusted based on application of second simulated camera parameters.

14. The computer readable storage medium of claim 10 , wherein the generated background scene includes a 2-Dimensional (2D) background scene randomly selected from a database of background scenes.

Assignees

• Tahoe Res Ltd

Inventors

• Bleiweiss Amit
• Paz Chen
• Levy Ofir
• Ben-Ari Itamar
• Yanai Yaron

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• G06T3/20

  Linear translation of whole images or parts thereof, e.g. panning · CPC title

• G06T15/20

  Perspective computation · CPC title

• G06T15/50

  Lighting effects · CPC title

• G06V20/64

  Three-dimensional [3D] objects · CPC title