Padding method for a convolutional neural network
AssigneeRobert Bosch GmbH
InventorsGeza Velkey, Kornel Istvan Kis, Peter Korosi-Szabo
Life of this patent
Estimating the term…
Abstract
A padding method for a convolutional neural network, in particular for concentrically configured data. The method includes receiving concentrically configured data of an object, the concentrically configured data correlating with an image, which was recorded concentrically to the object, deconvolving the concentrically configured data to form a data array, including real-coherent data on opposite sides of the data array, carrying out a convolution operation by using ring padding, in the case of ring padding, the real-coherent data of one side of the data array being utilized for padding the real-coherent data of a side of the data array opposite thereto, and/or vice versa.
Claims11
A device configured to control an at least semi-autonomous robot, the device configured to:
- receive surroundings data of the at least semi-autonomous robot, the surroundings data encompassing concentrically configured data which correlate with an image which was recorded concentrically to the at least semi-autonomous robot;
- apply an image recognition onto the recorded image including carrying out a padding method, the padding method including deconvolving the concentrically configured data to form a data array, including real-coherent data on opposite sides of the data array, and implementing a convolution operation by using ring padding, in the ring padding, the real-coherent data of one side of the data array being used for padding the real-coherent data of an opposite side of the data array;
- and
- control the at least semi-autonomous robot depending on the image recognition applied onto the recorded image.
A control method for an at least semi-autonomous robot, comprising the following steps:
- receiving surroundings data of the at least semi-autonomous robot, the surroundings data encompassing concentrically configured data which correlate with an image which was recorded concentrically to the at least semi-autonomous robot;
- applying an image recognition onto the recorded image including carrying out a padding method, the padding method including deconvolving the concentrically configured data to form a data array, including real-coherent data on opposite sides of the data array, and implementing a convolution operation by using ring padding, in the ring padding, the real-coherent data of one side of the data array being used for padding the real-coherent data of an opposite side of the data array;
- and
- controlling the at least semi-autonomous robot depending on the image recognition applied onto the recorded image.
A non-transitory machine-readable memory medium on which is stored a computer program for performing a control method for an at least semi-autonomous robot, the computer program, when executed by a computer, causing the computer to perform:
- receiving surroundings data of the at least semi-autonomous robot, the surroundings data encompassing concentrically configured data which correlate with an image which was recorded concentrically to the at least semi-autonomous robot;
- applying an image recognition onto the recorded image including carrying out a padding method, the padding method including deconvolving the concentrically configured data to form a data array, including real-coherent data on opposite sides of the data array, and implementing a convolution operation by using ring padding, in the ring padding, the real-coherent data of one side of the data array being used for padding the real-coherent data of an opposite side of the data array;
- and
- controlling the at least semi-autonomous robot depending on the image recognition applied onto the recorded image.
Dependencies are inferred from the claim text.