Training a classifier to detect open vehicle doors

Patent metadata

Field	Value
Publication number	US-11514310-B2
Application number	US-201816231297-A
Country	US
Kind code	B2
Filing date	Dec 21, 2018
Priority date	Dec 21, 2018
Publication date	Nov 29, 2022
Grant date	Nov 29, 2022

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a classifier to detect open vehicle doors. One of the methods includes obtaining a plurality of initial training examples, each initial training example comprising (i) a sensor sample from a collection of sensor samples and (ii) data classifying the sensor sample as characterizing a vehicle that has an open door; generating a plurality of additional training examples, comprising, for each initial training example: identifying, from the collection of sensor samples, one or more additional sensor samples that were captured less than a threshold amount of time before the sensor sample in the initial training example was captured; and training the machine learning classifier on first training data that includes the initial training examples and the additional training examples to generate updated weights for the machine learning classifier.

First claim

What is claimed is:

1. A computer-implemented method of training a machine learning classifier having a plurality of weights, wherein the machine learning classifier performs operations including:
receiving an input sensor sample that characterizes a first vehicle and is generated from sensor data captured by one or more sensors of a second vehicle, and
processing the input sensor sample to generate an open door score that represents a predicted likelihood that the first vehicle has an open door,
the computer-implemented method comprising:
obtaining a plurality of initial training examples, each initial training example comprising (i) a sensor sample from a collection of sensor samples and (ii) data classifying the sensor sample as characterizing a vehicle that has an open door;
generating a plurality of additional training examples, comprising, for each initial training example:
identifying, from the collection of sensor samples, one or more additional sensor samples that were captured less than a threshold amount of time before the sensor sample in the initial training example was captured,
classifying each additional sensor sample as a sensor sample that characterizes a vehicle that has an open door, and
generating one or more of the plurality of additional training examples, wherein each of generated additional training examples comprises: (i) one of the additional sensor samples that have been identified to have been captured less than the threshold amount of time before the sensor sample in the initial training example was captured and (ii) data classifying the additional sensor sample as characterizing a vehicle that has an open door; and

training the machine learning classifier on first training data that includes the initial training examples and the additional training examples to generate updated weights for the machine learning classifier to detect open vehicle doors.

2. The computer-implemented method of claim 1 , further comprising:
generating, using the machine learning classifier and in accordance with the updated weights, further training examples; and
training the machine learning classifier on second training data that includes the further training examples to generate further updated weights for the machine learning classifier.

3. The computer-implemented method of claim 2 , wherein training the machine learning classifier on second training data that includes the further training examples to generate further updated weights for the machine learning classifier comprises:
training the machine learning classifier on the second training data to generate further updated weights for the machine learning classifier starting from the updated weights for the machine learning classifier.

4. The computer-implemented method of claim 2 , wherein training the machine learning classifier on second training data that includes the further training examples to generate further updated weights for the machine learning classifier comprises:
training the machine learning classifier on the second training data to generate further updated weights for the machine learning classifier starting from initial weights for the machine learning classifier.

5. The computer-implemented method of claim 2 , wherein generating, using the machine learning classifier and in accordance with the updated weights, further training examples comprises:
processing each of a plurality of candidate sensor samples from the collection of sensor samples using the machine learning classifier and in accordance with the updated weights to generate a respective open door score for each candidate sensor sample; and
classifying each candidate sensor sample having an open door score that exceeds a threshold score as a sensor sample that characterizes a vehicle with an open door.

6. The method of claim 1 , wherein identifying, from the plurality of sensor samples, one or more additional sensor samples that were captured less than a threshold amount of time before the sensor sample in the initial training example was captured comprises:
identifying, as an additional sensor sample, each sensor sample in the plurality of samples that (i) characterizes the same vehicle as the sensor sample in the initial training example and (ii) was captured less than a threshold amount of time before the sensor sample in the initial training example was captured.

7. The computer-implemented method of claim 1 , wherein obtaining a plurality of initial training examples comprises:
identifying, from the collection of sensor samples, a plurality of candidate initial sensor samples wherein each candidate initial sensor sample includes a more than a threshold amount of measurements outside of an outline of a body of the vehicle characterized by the candidate initial sensor sample.

8. A system comprising one or more computers and one or more storage devices storing instructions that when executed by the one or more computers cause the one or more computers to perform operations for training a machine learning classifier having a plurality of weights, wherein the machine learning classifier performs operations including:
receiving an input sensor sample that characterizes a first vehicle and is generated from sensor data captured by one or more sensors of a second vehicle, and
processing the input sensor sample to generate an open door score that represents a predicted likelihood that the first vehicle has an open door,
the one or more computers perform operations including:
obtaining a plurality of initial training examples, each initial training example comprising (i) a sensor sample from a collection of sensor samples and (ii) data classifying the sensor sample as characterizing a vehicle that has an open door;
generating a plurality of additional training examples, comprising, for each initial training example:
identifying, from the collection of sensor samples, one or more additional sensor samples that were captured less than a threshold amount of time before the sensor sample in the initial training example was captured,
classifying each additional sensor sample as a sensor sample that characterizes a vehicle that has an open door, and
generating one or more of the plurality of additional training examples, wherein each of generated additional training examples comprises: (i) one of the additional sensor samples that have been identified to have been captured less than the threshold amount of time before the sensor sample in the initial training example was captured and (ii) data classifying the additional sensor sample as characterizing a vehicle that has an open door; and

training the machine learning classifier on first training data that includes the initial training examples and the additional training examples to generate updated weights for the machine learning classifier to detect open vehicle doors.

9. The system of claim 8 , the operations further comprising:
generating, using the machine learning classifier and in accordance with the updated weights, further training examples; and
training the machine learning classifier on second training data that includes the further training examples to generate further updated weights for the machine learning classifier.

10. The system of claim 9 , wherein training the machine learning classifier on second training data that includes the further training examples to generate further updated weights for the machine learning classifier comprises:
training the machine learning classifier on the second training data to generate further updated weights for the machine learning classifier starting from the updated weights for the machine learning classifier.

11. The system o

Assignees

• Waymo Llc

Inventors

• Mao Junhua
• Tsui Lo Po
• Li Congcong
• Walker Jr Edward Stephen

Classifications

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title

• G06V10/774

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06V10/776

  Validation; Performance evaluation · CPC title

• G06N3/08 Primary

  Learning methods · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title