Information processing apparatus, information processing method, and non-transitory computer-readable storage medium

Patent metadata

Field	Value
Publication number	US-12591983-B2
Application number	US-202318338520-A
Country	US
Kind code	B2
Filing date	Jun 21, 2023
Priority date	Jul 8, 2022
Publication date	Mar 31, 2026
Grant date	Mar 31, 2026

Abstract

An information processing apparatus comprises a first computation unit configured to obtain first features of an image of a tracking target, a second computation unit configured to obtain second features of an image of a search region, a third computation unit configured to obtain an inference tensor representing likelihoods that the tracking target is present at respective positions of the image of the search region, using the first features and the second features, and a fourth computation unit configured to obtain an inference map representing a position of the tracking target in the image of the search region, using the inference tensor.

First claim

What is claimed is:

1 . An information processing apparatus comprising:
at least one processor; and at least one memory storing instructions, which when executed by the processor, cause the information processing apparatus to: obtain first features of an image of a tracking target; obtain second features of an image of a search region; obtain an inference tensor representing likelihoods that the tracking target is present at respective positions of the image of the search region, using the first features and the second features; and obtain an inference map representing a position of the tracking target in the image of the search region, using the inference tensor, wherein the inference map representing the position of the tracking target in the image of the search region is obtained, using third features of the image of the tracking target and the inference tensor.

2 . The information processing apparatus according to claim 1 , wherein the instructions, when executed by the processor, further cause the information processing apparatus to:
reshape the first features into features whose number of dimensions is that of the inference tensor.

3 . The information processing apparatus according to claim 2 , wherein the apparatus further includes:
convolutional neural networks (CNNs), and wherein the CNN obtains the inference tensor by inputting the second features to the CNN in which the first features reshaped are set as weight parameters and performing computation of the CNN, and wherein the CNN is fully convolutional.

4 . The information processing apparatus according to claim 3 , wherein the first features and the second features are obtained by different CNNs.

5 . The information processing apparatus according to claim 3 , wherein the first features and the second features are obtained by the same CNN.

6 . The information processing apparatus according to claim 1 , wherein the instructions, when executed by the processor, further cause the information processing apparatus to:
reshape the first features by combining the first features with features held in advance.

7 . The information processing apparatus according to claim 1 , wherein the instructions, when executed by the processor, further cause the information processing apparatus to:
reshape features obtained based on the first features and the first features previously obtained by the first computation unit.

8 . The information processing apparatus according to claim 1 , the instructions, when executed by the processor, further cause the information processing apparatus to:
obtain an inference tensor representing likelihoods that a training target is present at respective position of an image of a search region, using first features of an image of the training target and second features of the image of the search region; and obtain an inference map representing a position of the training target in the image of the search region, using the inference tensor, wherein the inference map is used for training.

9 . The information processing apparatus according to claim 1 , the instructions, when executed by the processor, further cause the information processing apparatus to:
obtain an inference tensor representing likelihoods that a training target is present at respective position of an image of a search region, using first features of an image in which an image of the training target and an image of a target different from the training target are concatenated and second features of the image of the search region; and obtain an inference map representing a position of the training target in the image of the search region, using the inference tensor, wherein the inference map is used for training.

10 . An information processing method to be performed by an information processing apparatus, the method comprising:
obtaining first features of an image of a tracking target; obtaining second features of an image of a search region; obtaining an inference tensor representing likelihoods that the tracking target is present at respective positions of the image of the search region, using the first features and the second features; and obtaining an inference map representing a position of the tracking target in the image of the search region, using third features of the image of the tracking target and the inference tensor.

11 . The information processing method according to claim 10 , wherein the inference tensor is obtained by inputting the second features to the CNN in which the first features reshaped are set as weight parameters and performing computation of the CNN, and wherein the CNN is fully convolutional.

12 . A non-transitory computer-readable storage medium storing a computer program causing a computer to function as:
at least one processor; and at least one memory storing instructions, which when executed by the processor, cause the information processing apparatus to: obtain first features of an image of a tracking target; obtain second features of an image of a search region; obtain an inference tensor representing likelihoods that the tracking target is present at respective positions of the image of the search region, using the first features and the second features; and obtain an inference map representing a position of the tracking target in the image of the search region, using the inference tensor, wherein the inference map representing the position of the tracking target in the image of the search region is obtained, using third features of the image of the tracking target and the inference tensor.

13 . The non-transitory computer-readable storage medium according to claim 12 , wherein the inference tensor is obtained by inputting the second features to the CNN in which the first features reshaped are set as weight parameters and performing computation of the CNN, and wherein the CNN is fully convolutional.

Assignees

• Canon Kk

Inventors

• Takagi Yutaka
• Tate Shunta

Classifications

• G06V10/44

  Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components · CPC title

• G06V10/806 Primary

  of extracted features · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06T7/73

  using feature-based methods · CPC title

• G06T2207/20081

  Training; Learning · CPC title