Object detection improvement based on autonomously selected training samples

Patent metadata

Field	Value
Publication number	US-11610080-B2
Application number	US-202016854768-A
Country	US
Kind code	B2
Filing date	Apr 21, 2020
Priority date	Apr 21, 2020
Publication date	Mar 21, 2023
Grant date	Mar 21, 2023

Abstract

A method for generating positive and negative training samples is presented. The method includes identifying false positive images of an object based on multiple images of an environment. The method also includes generating positive training samples from a set of images of the object. The method further includes generating a negative training sample from the false positive image. The method still further includes training an object detection system based on the positive training samples and the negative training sample.

First claim

What is claimed is:

1. A method for generating positive and negative training samples, comprising:
identifying a set of objects in an environment based on performing a density clustering analysis on a plurality of images of the environment;
identifying false positive images of each object of the set of objects in an environment based on the plurality of images of an environment;
generating positive training samples from a respective set of images of each object of the set of objects;
generating a negative training sample from the false positive images; and
training an object detection system to detect each object in the set of objects based on the positive training samples and the negative training sample.

2. The method of claim 1 , further comprising estimating a true spatial location of each of the plurality of objects in the environment based on the density clustering analysis.

3. The method of claim 2 , further comprising calculating a three-dimensional mean of clusters to estimate the true spatial location of each of the plurality of objects.

4. The method of claim 1 , in which the set of images vary in at least one of distance, angle, or a combination thereof.

5. The method of claim 1 , in which the environment is a static environment.

6. An apparatus for generating positive and negative training samples, the apparatus comprising:
a processor; and
a memory coupled with the processor and storing instructions operable, when executed by the processor, to cause the apparatus:
to identify a set of objects in an environment based on performing a density clustering analysis on a plurality of images of the environment;
to identify false positive images of each object of the set of objects in an environment based on the plurality of images of an environment;
to generate positive training samples from a respective set of images of each object of the set of objects;
to generate a negative training sample from the false positive images; and
to train an object detection system to detect each object in the set of objects based on the positive training samples and the negative training sample.

7. The apparatus of claim 6 , in which execution of the instructions further cause the apparatus to estimate a true spatial location of each of the plurality of objects in the environment based on the density clustering analysis.

8. The apparatus of claim 7 , in which execution of the instructions further cause the apparatus to calculate a three-dimensional mean of clusters to estimate the true spatial location of each of the plurality of objects.

9. The apparatus of claim 6 , in which the set of images vary in at least one of distance, angle, or a combination thereof.

10. The apparatus of claim 6 , in which the environment is a static environment.

11. A non-transitory computer-readable medium having program code recorded thereon for generating positive and negative training samples, the program code executed by a processor and comprising:
program code to identify a set of objects in an environment based on performing a density clustering analysis on a plurality of images of the environment;
program code to identify false positive images of each object of the set of objects in an environment based on the plurality of images of an environment;
program code to generate positive training samples from a respective set of images of each object of the set of objects;
program code to generate a negative training sample from the false positive images; and
program code to train an object detection system to detect each object in the set of objects based on the positive training samples and the negative training sample.

12. The non-transitory computer-readable medium of claim 11 , further comprising program code to estimate a true spatial location of each of the plurality of objects in the environment based on the density clustering analysis.

13. The non-transitory computer-readable medium of claim 12 , further comprising program code to calculate a three-dimensional mean of clusters to estimate the true spatial location of each of the plurality of objects.

14. The non-transitory computer-readable medium of claim 11 , in which the set of images vary in at least one of distance, angle, or a combination thereof.

Assignees

• Toyota Res Inst Inc

Inventors

• Northcutt Brandon
• Bouma Katarina
• Stone Kevin
• Mushegian Konstantine

Classifications

• H03M7/16

  Conversion to or from unit-distance codes, e.g. Gray code, reflected binary code · CPC title

• H03M1/56

  Input signal compared with linear ramp · CPC title

• H03K19/21

  EXCLUSIVE-OR circuits, i.e. giving output if input signal exists at only one input; COINCIDENCE circuits, i.e. giving output only if all input signals are identical · CPC title

• H03K19/20

  characterised by logic function, e.g. AND, OR, NOR, NOT circuits (H03K19/003 - H03K19/01 take precedence) · CPC title

• H04N25/75 Primary

  Circuitry for providing, modifying or processing image signals from the pixel array · CPC title