US 12,456,051 B2
Kernel transfer
Kwun Ho Ngan, Cambridge (GB); Artur Garcez, Morden (GB); and Joseph Townsend, High Wycombe (GB)
Assigned to Fujitsu Limited, Kawasaki (JP)
Filed by Fujitsu Limited, Kawasaki (JP)
Filed on Dec. 15, 2022, as Appl. No. 18/081,719.
Claims priority of application No. 22153860 (EP), filed on Jan. 28, 2022.
Prior Publication US 2023/0281441 A1, Sep. 7, 2023
Int. Cl. G06N 3/08 (2023.01); G06N 3/045 (2023.01); G06V 10/762 (2022.01); G06V 10/82 (2022.01)
CPC G06N 3/08 (2013.01) [G06N 3/045 (2023.01); G06V 10/762 (2022.01); G06V 10/82 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
for each of a plurality of input samples of a first domain, obtaining an output from each of a plurality of kernels in an extraction layer of a first trained convolutional neural network, wherein the first convolutional neural network is configured to identify one or more features in an image;
generating a plurality of aggregate maps corresponding respectively to at least some of the kernels by, for each of the at least some of the kernels, aggregating the outputs from the kernel and corresponding to at least some of the input samples to generate an aggregate map corresponding to that kernel;
resizing the aggregate maps to a lower resolution to generate a plurality of region maps corresponding to the aggregate maps, respectively;
clustering the region maps to generate clusters of region maps, each cluster comprising region maps having similar regions; and
training, using input samples of a second domain, a second convolutional neural network with a kernel weight of at least one of the kernels which corresponds to at least one of the image regions of at least one of the clusters.
 
19. A computer program which, when run on a computer, causes the computer to carry out a method comprising:
for each of a plurality of input samples of a first domain, obtaining an output from each of a plurality of kernels in an extraction layer of a first trained convolutional neural network, wherein the first convolutional neural network is configured to identify one or more features in an image;
generating a plurality of aggregate maps corresponding respectively to at least some of the kernels by, for each of the at least some of the kernels, aggregating the outputs from the kernel and corresponding to at least some of the input samples to generate an aggregate map corresponding to that kernel;
resizing the aggregate maps to a lower resolution to generate a plurality of region maps corresponding to the aggregate maps, respectively;
clustering the region maps to generate clusters of region maps, each cluster comprising region maps having similar regions; and
training, using input samples of a second domain, a second convolutional neural network with a kernel weight of at least one of the kernels which corresponds to at least one of the image regions of at least one of the clusters.
 
20. An information processing apparatus comprising a memory and a processor connected to the memory, wherein the processor is configured to:
for each of a plurality of input samples of a first domain, obtain an output from each of a plurality of kernels in an extraction layer of a first trained convolutional neural network, wherein the first convolutional neural network is configured to identify one or more features in an image;
generate a plurality of aggregate maps corresponding respectively to at least some of the kernels by, for each of the at least some of the kernels, aggregating the outputs from the kernel and corresponding to at least some of the input samples to generate an aggregate map corresponding to that kernel;
resize the aggregate maps to a lower resolution to generate a plurality of region maps corresponding to the aggregate maps, respectively;
cluster the region maps to generate clusters of region maps, each cluster comprising region maps having similar regions; and
train, using input samples of a second domain, a second convolutional neural network with a kernel weight of at least one of the kernels which corresponds to at least one of the image regions of at least one of the clusters.