Methods and Systems for Counting People

Patent metadata

Field	Value
Publication number	US-2016358074-A1
Application number	US-201615163833-A
Country	US
Kind code	A1
Filing date	May 25, 2016
Priority date	Jun 5, 2015
Publication date	Dec 8, 2016
Grant date	—

Abstract

Various systems and methods for counting people. For example, one method involves receiving input data at an analytics system that includes a neural network. The input data includes a representation of an environment, including representations of several people. The method also includes identifying the representations of the people in the representation of the environment. The method also includes updating an output value that indicates the number of people identified as being present in the environment.

First claim

What is claimed is:

1 . A method comprising:
receiving input data at a processor of an analytics system, wherein
the input data comprises a representation of an environment at a first point in time,
the representation of the environment comprises a plurality of representations of persons, and
the analytics system comprises an artificial neural network;
identifying, using the processor, a representation of a person of the plurality of representations; and updating an output value, wherein
the output value indicates a number of persons identified as being present within the input data.

2 . The method of claim 1 , further comprising:
selecting an output routine based on the output value; and performing the output routine.

3 . The method of claim 1 , further comprising:
automatically compensating for a plurality of occlusions, wherein
the input data comprises a visual representation of the environment, and
the visual representation comprises the plurality of occlusions.

4 . The method of claim 1 , further comprising:
converging sensor information from a plurality of input sources to form the input data, wherein
the sensor information comprises image data and location information received from one or more wireless devices.

5 . The method of claim 1 , wherein
the identifying comprises generating a feature vector using a generative process, and the feature vector comprises a sparse distributed representation of the input data.

6 . The method of claim 5 , wherein
the generating comprises performing a discriminative process on the feature vector.

7 . The method of claim 1 , further comprising:
generating an augmented data set, wherein
the generating the augmented data set comprises injecting one or more modifications into a ground truth data set, and
the one or more modifications are selected by a domain expert.

8 . The method of claim 7 , further comprising:
generating a semantic fingerprint that comprises a probability distribution for the output value

9 . The method of claim 8 , further comprising:
performing one or more pre-training operations of the analytics system, wherein
the pre-training operations utilize the augmented data set and the semantic fingerprint.

10 . A system comprising:
a memory; and a processor coupled to the memory and configured to:
receive input data at the processor, wherein
the input data comprises a representation of an environment at a first point in time,
the representation of the environment comprises a plurality of representations of persons, and
the system comprises an artificial neural network;

identify, using the processor, a representation of a person of the plurality of representations; and
update an output value, wherein
the output value indicates a number of persons identified as being present within the input data.

11 . The system of claim 10 , wherein the processor is further configured to:
select an output routine based on the output value; and perform the output routine.

12 . The system of claim 10 , wherein the processor is further configured to:
automatically compensate for a plurality of occlusions, wherein
the input data comprises a visual representation of the environment, and
the visual representation comprises the plurality of occlusions.

13 . The system of claim 10 , wherein the processor is further configured to:
converge sensor information from a plurality of input sources to form the input data, wherein
the sensor information comprises image data and location information received from one or more wireless devices.

14 . The system of claim 10 , wherein
identifying the representation of the person comprises generating a feature vector using a generative process, and the feature vector comprises a sparse distributed representation of the input data.

15 . The system of claim 14 , wherein
the generating comprises performing a discriminative process on the feature vector.

16 . The system of claim 10 , wherein the processor is further configured to:
generate an augmented data set, wherein
generating the augmented data set comprises injecting one or more modifications into a ground truth data set, and
the one or more modifications are selected by a domain expert.

17 . The system of claim 10 , wherein the processor is further configured to:
generate a semantic fingerprint that comprises a probability distribution for the output value

18 . The system of claim 10 , wherein the processor is further configured to:
perform one or more pre-training operations of the analytics system, wherein
the pre-training operations utilize the augmented data set and the semantic fingerprint.

19 . A non-transitory computer readable storage medium comprising program instructions executable to:
receive input data at a processor of an analytics system, wherein
the input data comprises a representation of an environment at a first point in time,
the representation of the environment comprises a plurality of representations of persons, and
the system comprises an artificial neural network;
identify, using the processor, a representation of a person of the plurality of representations; and update an output value, wherein
the output value indicates a number of persons identified as being present within the input data.

20 . The non-transitory computer readable storage medium of claim 19 , wherein the program instructions are further executable to:
select an output routine based on the output value; and perform the output routine.

Assignees

• Cisco Tech Inc

Inventors

• Latapie Hugo M
• Fenoglio Enzo
• Pandey Santosh G
• Surcouf Andre

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/082

  modifying the architecture, e.g. adding, deleting or silencing nodes or connections · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/0475

  Generative networks · CPC title

• G06N3/096

  Transfer learning · CPC title