Information processing method and apparatus, and image pickup device

• US 7,991,719 B2
• Filed: 06/02/2005
• Issued: 08/02/2011
• Est. Priority Date: 06/03/2004
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Mitarai, Yusuke, Matsugu, Masakazu, Mori, Katsuhiko, Ishii, Mie
• Primary Examiner(s): Gaffin; Jeffrey A
• Assistant Examiner(s): Wong; Lut
• Application Number:
  US11/579,981
• Publication Number:
  US 20070244842A1
• Time in Patent Office: 2,252 Days
• Field of Search: 706/28
• US Class Current: 706/28
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/082 modifying the architecture, e.g. adding, deleting or silencing nodes or connections
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 2207/30201 Face
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06V 10/454 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]

Abstract

An output value of neuron within an objective layer of a hierarchical neural network is computed. The data of the output value of neuron is stored in a memory only if the output value of neuron is greater than or equal to a predetermined value by referring to the computed output value of neuron within the objective layer. When the data of the output value of neuron on a former layer of objective layer is read from the memory, the data having a predetermined value is read, instead of the data of the output value of neuron not stored in the memory.

Claims

1. An information processing apparatus for detecting a predetermined pattern in an input image, comprising:

a hierarchical neural network having an input layer for inputting data and having a plurality of detection layers and integration layers arranged alternatively, wherein (i) a first detection layer is configured to detect and output one or more feature amounts from the data input to the input layer, (ii) each detection layer except for the first detection layer is configured to detect and output one or more feature amounts from the output of the integration layer at the former stage, and (iii) each integration layer is configured to integrate the outputted features from the detection layer at the former stage and output the result of integration;

an operation processing section that controls the plurality of detection layers and integrations layers to sequentially set each layer to be the current layer, compute an input sum for a neuron in the current layer using an output value of a neuron within a former layer that precedes the current layer and a weight coefficient between the current layer and the former layer, and compute an output value for the neuron in the current layer based on the input sum; and

a storage control unit that stores data of output values that are greater than or equal to a predetermined value in a memory, by comparing the output value of each neuron within a detection layer computed by said operation processing section to the predetermined value,

wherein said operation processing section performs control to store data of the output value of each neuron within an integration layer computed by said operation processing section in the memory,

wherein said operation processing section performs control to obtain the data of the output value of a neuron within the former layer that precedes the current layer by reading the data of the output value from the memory and performs control so that data having a predetermined non-zero value is used instead of the data of the output value of a neuron that is not stored in the memory,

wherein the input sum for a neuron is computed based on the formula;

2. The information processing apparatus according to claim 1, wherein said output value computation means omits the computation of an output value of neuron, when all the data to be referred to in computing the output value of neuron within the detection layer is not supplied by said supply means.

3. An image pickup device mounting the information processing apparatus according to claim 1.

4. An information processing method for detecting a predetermined pattern in an input image, comprising:

using a hierarchical neural network to perform the following steps, the hierarchical neural network having an input layer for inputting data and having a plurality of detection layers and integration layers arranged alternately, wherein (i) a first detection layer is configured to detect and output one or more feature amounts from the data input by the input layer, (ii) each detection layer except for the first detection layer is configured to detect and output one or more feature amounts from the output of the integration layer at the former stage, and (iii) the integration layer is configured to integrate the outputted features from the detection layer at the former stage and output the result of integration;

an output value computation step of sequentially setting each layer to be the current layer, computing an input sum for a neuron in the current layer using an output value of neuron within a former layer that precedes the current layer and a weight coefficient between the current layer and the former layer, and computing an output value for the neuron in the current layer based on the input sum;

a first storage control step of storing data of output values that are greater than or equal to a predetermined value in a memory, by comparing the output value of each neuron within a detection layer computed in the output value computation step to the predetermined value;

a second storage control step of storing data of the output value of each neuron within an integration layer computed in the output value computation step, in the memory; and

a supply step of supplying the data of the output value of the neuron within the former layer that precedes the current layer, which is referred to in the output value computation step, by reading the data of the output value from the memory and of supplying data having a predetermined non-zero value instead of the data of the output value of a neuron not stored in the memory,

wherein the input sum for a neuron is computed based on the formula;

5. A computer readable storage medium storing a program for causing an information processing apparatus to perform a method according to claim 4.