# US 11,908,142 B2 · Method, artificial neural network, device, computer program, and machine-readable memory medium for the semantic segmentation of image data

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

- Number: US 11,908,142 B2
- Document ID: US11908142B2
- Status: Granted
- Kind: B2
- Application number: 17/282,372
- Filed: 2019-09-26
- Published: 2021-11-04
- Granted: 2024-02-20
- Earliest claimed: 2018-10-05
- Assignees: Robert Bosch GmbH
- Applicants: Robert Bosch GmbH
- Inventors: Ferran Diego Andilla; Dimitrios Bariamis; Masato Takami; Uwe Brosch
- Web page: https://patentgrep.com/patent/US11908142B2

## CPC

- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06T 7/11: Region-based segmentation
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/09: Supervised learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

A method for the computing and memory resource-conserving semantic segmentation of image data of an imaging sensor with the aid of an artificial neural network, in particular, a convolutional neural network, the artificial neural network including an encoder path and a decoder path. The method includes: dividing an input tensor into at least one first slice tensor and at least one second slice tensor as a function of a division function, the input tensor being dependent on the image data; outputting the at least one first slice tensor to the decoder path of the neural network; connecting the at least one first slice tensor to the at least one second slice tensor as a function of a connecting function to obtain an output tensor; and outputting the output tensor to the encoder path of the artificial neural network.

## Claims

### Claim 1

A method for computing and memory resource-conserving semantic segmentation of image data of an imaging sensor using an artificial neural network, the artificial neural network including an encoder path and a decoder path, comprising the following steps:
dividing an input tensor into at least one first slice tensor and at least one second slice tensor as a function of a division function, the input tensor being dependent on the image data;
outputting the at least one first slice tensor to the decoder path of the artificial neural network;
connecting the at least one first slice tensor to the at least one second slice tensor as a function of a connecting function to obtain an output tensor; and
outputting the output tensor to the encoder path of the neural network.

### Claim 2 (depends on 1)

The method as recited in claim 1, wherein the artificial neural network is a convolutional neural network.

### Claim 3 (depends on 1)

The method as recited in claim 1, wherein, in the step of dividing, the division function is configured in such a way that only a subset of feature maps of the input tensor is selected for forming the at least one first slice tensor.

### Claim 4 (depends on 1)

The method as recited in claim 1, wherein the connecting function is configured in such a way that a dimension of the input tensor is maintained.

### Claim 5 (depends on 1)

The method as recited in claim 1, further comprising a preceding step of receiving, the input tensor and the division function being received in the step of receiving.

### Claim 6 (depends on 1)

The method as recited in claim 1, wherein, in the step of dividing, a first function of a neural network is applied to the at least one first slice tensor, and a second function of a neural network is applied to the at least one second slice tensor.

### Claim 7 (depends on 6)

The method as recited in claim 6, wherein the division function is configured in such a way that it includes the number of feature maps to be computed and respective functions of an artificial neural network for computing the at least one first slice tensor and the at least one second slice tensor.

### Claim 8

An artificial neural network, for the semantic segmentation of image data of an imaging sensor, the artificial neural network including an encoder path for classification of the image data and a decoder path for localization of the image data, the artificial neural network configured to:
divide an input tensor into at least one first slice tensor and at least one second slice tensor as a function of a division function, the input tensor being dependent on the image data;
output the at least one first slice tensor to the decoder path of the artificial neural network;
connect the at least one first slice tensor to the at least one second slice tensor as a function of a connecting function to obtain an output tensor; and
output the output tensor to the encoder path of the neural network.

### Claim 9 (depends on 8)

The artificial neural network as recited in claim 8, wherein the artificial neural network is a convolutional neural network.

### Claim 10

A device including an artificial neural network for the semantic segmentation of image data of an imaging sensor, the artificial neural network including an encoder path for classification of the image data and a decoder path localization of the image data, the device configured to:
divide an input tensor into at least one first slice tensor and at least one second slice tensor as a function of a division function, the input tensor being dependent on the image data;
output the at least one first slice tensor to the decoder path of the artificial neural network;
connect the at least one first slice tensor to the at least one second slice tensor as a function of a connecting function to obtain an output tensor; and
output the output tensor to the encoder path of the neural network.

### Claim 11

A non-transitory machine-readable memory medium on which is stored a computer program for computing and memory resource-conserving semantic segmentation of image data of an imaging sensor using an artificial neural network, the artificial neural network including an encoder path and a decoder path, the computer program, when executed by a computer, causing the computer to perform the following steps:
dividing an input tensor into at least one first slice tensor and at least one second slice tensor as a function of a division function, the input tensor being dependent on the image data;
outputting the at least one first slice tensor to the decoder path of the artificial neural network;
connecting the at least one first slice tensor to the at least one second slice tensor as a function of a connecting function to obtain an output tensor; and
outputting the output tensor to the encoder path of the neural network.