DOMAIN ADAPTATION AND FUSION USING WEAKLY SUPERVISED TARGET-IRRELEVANT DATA
Aspects include receiving a request to perform an image classification task in a target domain. The image classification task includes identifying a feature in images in the target domain. Classification information related to the feature is transferred from a source domain to the target domain. The transferring includes receiving a plurality of pairs of task-irrelevant images that each includes a task-irrelevant image in the source domain and in the target domain. The task-irrelevant image in the source domain has a fixed correspondence to the task-irrelevant image in the target domain. A target neural network is trained to perform the image classification task in the target domain. The training is based on the plurality of pairs of task-irrelevant images. The image classification task is performed in the target domain and includes applying the target neural network to an image in the target domain and outputting an identified feature.
1 . A method comprising:
receiving, by a system comprising one or more processors, a request to perform an image classification task in a target domain, the image classification task including identifying a feature in images in the target domain;
transferring classification information related to the feature from a source domain to the target domain, the transferring comprising:
receiving a plurality of pairs of task-irrelevant images, each image pair comprising a task-irrelevant image in the source domain and a task-irrelevant image in the target domain, the task-irrelevant image in the source domain having a fixed correspondence to the task-irrelevant image in the target domain; and
training a target neural network to perform the image classification task in the target domain, the training based on the plurality of pairs of task-irrelevant images; and
performing the image classification task in the target domain, the performing including applying the target neural network to an image in the target domain and outputting an identified feature.
2 . The method of claim 1 , wherein the fixed correspondence is a spatial relation.
3 . The method of claim 1 , wherein the classification information related to the feature from the source domain is contained in a source neural network.
4 . The method of claim 3 , further comprising training the source neural network based at least in part on labeled images in the source domain and concurrently with the training of the training of the target neural network.
5 . The method of claim 4 , wherein the labeled images in the source domain include task-relevant images in the source domain.
6 . The method of claim 5 , wherein labeled images in the source domain include simulated data.
7 . The method of claim 1 , further comprising training a joint neural network using only the pairs of task-irrelevant images and task-relevant labeled data in the source domain.
8 . The method of claim 1 , wherein the request is received from an integrated perception system (IPS) that monitors a geographic location and the identified feature is output to the IPS.
9 . A system comprising:
a memory having computer readable instructions; and
one or more processors for executing the computer readable instructions, the computer readable instructions controlling the one or more processors to perform operations comprising:
receiving a request to perform an image classification task in a target domain, the image classification task including identifying a feature in images in the target domain;
transferring classification information related to the feature from a source domain to the target domain, the transferring comprising:
receiving a plurality of pairs of task-irrelevant images, each image pair comprising a task-irrelevant image in the source domain and a task-irrelevant image in the target domain, the task-irrelevant image in the source domain having a fixed correspondence to the task-irrelevant image in the target domain; and
training a target neural network to perform the image classification task in the target domain, the training based on the plurality of pairs of task-irrelevant images; and
performing the image classification task in the target domain, the performing including applying the target neural network to an image in the target domain and outputting an identified feature.
10 . The system of claim 9 , wherein the fixed correspondence is a spatial relation.
11 . The system of claim 9 , wherein the classification information related to the feature from the source domain is contained in a source neural network.
12 . The system of claim 11 , wherein the operations further comprise training the source neural network based at least in part on labeled images in the source domain and concurrently with the training of the training of the target neural network.
13 . The system of claim 12 , wherein the labeled images in the source domain include task-relevant images in the source domain.
14 . The system of claim 13 , wherein labeled images in the source domain include simulated data.
15 . The system of claim 9 , wherein the operations further comprise training a joint neural network using only the pairs of task-irrelevant images and task-relevant labeled data in the source domain.
16 . The system of claim 9 , wherein the request is received from an integrated perception system (IPS) that monitors a geographic location and the identified feature is output to the IPS.
17 . A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform operations comprising:
receiving a request to perform an image classification task in a target domain, the image classification task including identifying a feature in images in the target domain;
transferring classification information related to the feature from a source domain to the target domain, the transferring comprising:
receiving a plurality of pairs of task-irrelevant images, each image pair comprising a task-irrelevant image in the source domain and a task-irrelevant image in the target domain, the task-irrelevant image in the source domain having a fixed correspondence to the task-irrelevant image in the target domain; and
training a target neural network to perform the image classification task in the target domain, the training based on the plurality of pairs of task-irrelevant images; and
performing the image classification task in the target domain, the performing including applying the target neural network to an image in the target domain and outputting an identified feature.
18 . The computer program product of claim 17 , wherein the classification information related to the feature from the source domain is contained in a source neural network.
19 . The computer program product of claim 17 , wherein the operations further comprise training the source neural network based at least in part on labeled images in the source domain and concurrently with the training of the training of the target neural network.
20 . The computer program product of claim 17 , wherein the operations further comprise training a joint neural network using only the pairs of task-irrelevant images and task-relevant labeled data in the source domain.