# US 2021/0232856 A1 · Image processing method, apparatus, electronic device and storage medium

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0232856 A1
- Document ID: US20210232856A1
- Status: Published application
- Kind: A1
- Application number: 17/213,746
- Filed: 2021-03-26
- Published: 2021-07-29
- Earliest claimed: 2020-04-17
- Assignees: Beijing Baidu Netcom Science Technology Co., Ltd.
- Applicants: Beijing Baidu Netcom Science and Technology Co., Ltd.
- Inventors: Yingying Li; Xiao Tan; Minyue Jiang; Hao Sun
- Web page: https://patentgrep.com/patent/US20210232856A1

## CPC

- G06V 10/806: {of extracted features}
- G06F 18/211: Selection of the most significant subset of features
- G06F 18/253: {of extracted features}
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/09: Supervised learning
- G06T 3/00: Geometric image transformations in the plane of the image
- G06V 10/7715: {Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods}
- G06V 10/82: using neural networks
- G06N 3/08: Learning methods
- Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE

## Abstract

The present disclosure provides an image processing method. An image to be classified is input into a feature extraction model to generate N dimensional features. Dimension fusion is performed on M features of the N dimensional features to obtain M dimension fusion features. The image to be classified is processed based on M dimension fusion features and remaining features of the N dimensional features other than the M features.

## Claims

### Claim 1

An image processing method, comprising:
inputting an image into a feature extraction model to generate N dimensional features, N being a positive integer;
performing dimension fusion on M features selected from the N dimensional features to obtain M dimension fusion features, M being a positive integer and less than or equal to N; and
processing the image based on M dimension fusion features and remaining features of the N dimensional features other than the M features.

### Claim 2 (depends on 1)

The method according to claim 1, wherein performing the dimension fusion on the M features selected from the N dimensional features to obtain the M dimension fusion features comprises:
selecting the M features from the N dimensional features based on the number of dimensions corresponding to the N dimensional features; and
performing the dimension fusion on the M features to obtain the M dimension fusion features.

### Claim 3 (depends on 2)

The method according to claim 2, wherein the selecting the M features from the N dimensional features based on the number of dimensions corresponding to the N dimensional features comprises:
selecting the M features from a first dimension range and a second dimension range of the N dimensional features, the second dimension range being larger than the first dimension range.

### Claim 4 (depends on 2)

The method according to claim 2, wherein performing the dimension fusion on the M features selected from the N dimensional features to obtain the M dimension fusion features comprises:
transforming the M features to generate M dimensional features; and
convoluting the M dimensional features to obtain the M dimension fusion features.

### Claim 5 (depends on 4)

The method according to claim 4, wherein convoluting the M dimensional features to obtain the M dimension fusion features comprises:
S1, convoluting each of the M dimensional features for two times based on a preset channel and a preset convolution kernel value to obtain a first convolution result and a second convolution result respectively;
S2, fusing the first convolution result and the second convolution result to generate an ith fusion result, where i=1;
S3: convoluting the ith fusion result based on the preset channel and the preset convolution kernel value to generate an (i+2)th convolution result;
S4, fusing the first convolution result, the second convolution result . . . until the (i+2)th convolution result to generate an (i+1)th fusion result;
S5, updating i to i=i+1, and returning to perform S3 and S4, in a case where i is less than a value obtained by subtracting 2 from the preset convolution kernel value; and
S6: determining the (i+1)th fusion result as the M dimension fusion features in a case where i is equal to a value obtained by subtracting 2 from the preset convolution kernel value.

### Claim 6 (depends on 1)

The method according to claim 1, wherein processing the image comprises image classifying, image dividing or image detecting.

### Claim 7

An electronic device, comprising:
at least one processor; and
a memory communicatively connected with the at least one processor; wherein,
the memory is configured to store instructions executable by the at least one processor, when the instructions are executed by the at least one processor, the at least one processor is configured to:
input an image into a feature extraction model to generate N dimensional features, N being a positive integer;
perform dimension fusion on M features selected from the N dimensional features to obtain M dimension fusion features, M being a positive integer and less than or equal to N; and
process the image based on M dimension fusion features and remaining features of the N dimensional features other than the M features.

### Claim 8 (depends on 7)

The electronic device according to claim 7, wherein the at least one processor is further configured to:
select the M features from the N dimensional features based on the number of dimensions corresponding to the N dimensional features; and
perform the dimension fusion on the M features to obtain the M dimension fusion features.

### Claim 9 (depends on 8)

The electronic device according to claim 8, wherein the at least one processor is further configured to:
select the M features from a first dimension range and a second dimension range of the N dimensional features, the second dimension range being larger than the first dimension range.

### Claim 10 (depends on 8)

The electronic device according to claim 8, wherein the at least one processor is further configured to:
transform the M features to generate M dimensional features; and
convolute the M dimensional features to obtain the M dimension fusion features.

### Claim 11 (depends on 10)

The electronic device according to claim 10, wherein the at least one processor is further configured to:
S1, convolute each of the M dimensional features for two times based on a preset channel and a preset convolution kernel value to obtain a first convolution result and a second convolution result respectively;
S2, fuse the first convolution result and the second convolution result to generate an ith fusion result, where i=1;
S3: convolute the ith fusion result based on the preset channel and the preset convolution kernel value to generate an (i+2)th convolution result;
S4, fuse the first convolution result, the second convolution result . . . until the (i+2)th convolution result to generate an (i+1)th fusion result;
S5, update i to i=i+1, and returning to perform S3 and S4, in a case where i is less than a value obtained by subtracting 2 from the preset convolution kernel value; and
S6: determine the (i+1)th fusion result as the M dimension fusion features in a case where i is equal to a value obtained by subtracting 2 from the preset convolution kernel value.

### Claim 12 (depends on 7)

The electronic device according to claim 7, wherein processing the image comprises image classifying, image dividing or image detecting.

### Claim 13

A non-transitory computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions are used to make the computer to execute an image processing method, the method comprising:
inputting an image into a feature extraction model to generate N dimensional features, N being a positive integer;
performing dimension fusion on M features selected from the N dimensional features to obtain M dimension fusion features, M being a positive integer and less than or equal to N; and
processing the image based on M dimension fusion features and remaining features of the N dimensional features other than the M features.

### Claim 14 (depends on 13)

The non-transitory computer-readable storage medium according to claim 13, wherein performing the dimension fusion on the M features selected from the N dimensional features to obtain the M dimension fusion features comprises:
selecting the M features from the N dimensional features based on the number of dimensions corresponding to the N dimensional features; and
performing the dimension fusion on the M features to obtain the M dimension fusion features.

### Claim 15 (depends on 14)

The non-transitory computer-readable storage medium according to claim 14, wherein the selecting the M features from the N dimensional features based on the number of dimensions corresponding to the N dimensional features comprises:
selecting the M features from a first dimension range and a second dimension range of the N dimensional features, the second dimension range being larger than the first dimension range.

### Claim 16 (depends on 14)

The non-transitory computer-readable storage medium according to claim 14, wherein performing the dimension fusion on the M features selected from the N dimensional features to obtain the M dimension fusion features comprises:
transforming the M features to generate M dimensional features; and
convoluting the M dimensional features to obtain the M dimension fusion features.

### Claim 17 (depends on 16)

The non-transitory computer-readable storage medium according to claim 16, wherein convoluting the M dimensional features to obtain the M dimension fusion features comprises:
S1, convoluting each of the M dimensional features for two times based on a preset channel and a preset convolution kernel value to obtain a first convolution result and a second convolution result respectively;
S2, fusing the first convolution result and the second convolution result to generate an ith fusion result, where i=1;
S3: convoluting the ith fusion result based on the preset channel and the preset convolution kernel value to generate an (i+2)th convolution result;
S4, fusing the first convolution result, the second convolution result . . . until the (i+2)th convolution result to generate an (i+1)th fusion result;
S5, updating i to i=i+1, and returning to perform S3 and S4, in a case where i is less than a value obtained by subtracting 2 from the preset convolution kernel value; and
S6: determining the (i+1)th fusion result as the M dimension fusion features in a case where i is equal to a value obtained by subtracting 2 from the preset convolution kernel value.

### Claim 18 (depends on 13)

The non-transitory computer-readable storage medium according to claim 13, wherein processing the image comprises image classifying, image dividing or image detecting.
