State determination device, state determination method and recording medium
A state determination device according to the present invention includes: a memory; and at least one processor coupled to the memory. The processor performs operations. The operations including: combining output from a plurality of determination models for an input image of a road, the plurality of determination models each learned using training data in which at least one of a state of a moving object mounting an imaging device that acquires an image of the road, a state of the road, and an external environment is different.
1. A state determination device comprising:
a memory storing instructions; and
at least one processor configured to execute the instructions to perform operations comprising:
determining weights for combining outputs of a plurality of determination models respectively determined by the plurality of determination models using a same input image of a target road, each of the plurality of determination models having been trained using training data in which at least one of a state of a moving object on which an imaging device that is to acquire an image of the road is mounted, a state of the road in the acquired image, and an external environment in which the image of the road has been acquired by the imaging device is different than that in the training data used to train one or more other of the plurality of determination models; and
combining the outputs of the plurality of determination models for the same input image of the target road using the determined weights.
2. The state determination device according to claim 1 , wherein the operations further comprise:
determining and outputting a state of the target road based on a result of combining the outputs of the plurality of determination models for the same input image.
3. The state determination device according to claim 1 , wherein combining the outputs of the plurality of determination models for the image of the target road using the determined weights comprises respectively multiplying outputs of the plurality of determination models by the determined weights.
4. The state determination device according to claim 3 , wherein the operations further comprise:
learning a parameter for determining the weights using training data for learning the weights.
5. The state determination device according to claim 4 , wherein learning the parameter for determining the eights using the training data for learning the weights comprises:
learning the parameter for determining the weights in accordance with the external environment.
6. The state determination device according to claim 1 , wherein the plurality of determination models respectively determine the outputs as a state of deterioration of the road and structures related to the road.
7. A state determination method performed by a computer and comprising:
determining weights for combining outputs of a plurality of determination models respectively determined by the plurality of determination models using a same input image of a target road, each of the plurality of determination models having been trained using training data in which at least one of a state of a moving object on which an imaging device that is to acquire an image of the road is mounted, a state of the road in the acquired image, and an external environment in which the image of the road has been acquired by the imaging device is different than that in the training data used to train one or more other of the plurality of determination models; and
combining the outputs of the plurality of determination models for the same input image of the target road using the determined weights.
8. The state determination method according to claim 7 , further comprising:
determining and outputting a state of the target road based on a result of combining the outputs of the plurality of determination models for the same input image.
9. The state determination method according to claim 7 , wherein combining the outputs of the plurality of determination models for the image of the target road using the determined weights comprises respectively multiplying outputs of the plurality of determination models by the determined weights.
10. The state determination method according to claim 9 , further comprising:
learning a parameter for determining the weights using training data for learning the weights.
11. The state determination method according to claim 10 , wherein learning the parameter for determining the eights using the training data for learning the weights comprises:
learning the parameter for determining the weights in accordance with the external environment.
12. The state determination method according to claim 7 , wherein the plurality of determination models respectively determine the outputs as a state of deterioration of the road and structures related to the road.
13. A non-transitory computer-readable recording medium storing a program executable by a computer to perform a method comprising:
determining weights for combining outputs of a plurality of determination models respectively determined by the plurality of determination models using a same input image of a target road, each of the plurality of determination models having been trained using training data in which at least one of a state of a moving object on which an imaging device that is to acquire an image of the road is mounted, a state of the road in the acquired image, and an external environment in which the image of the road has been acquired by the imaging device is different than that in the training data used to train one or more other of the plurality of determination models; and
combining the outputs of the plurality of determination models for the same input image of the target road using the determined weights.
14. The non-transitory computer-readable recording medium according to claim 13 , wherein the method further comprises:
determining and outputting a state of the target road based on a result of combining the outputs of the plurality of determination models for the same input image.
15. The non-transitory computer-readable recording medium according to claim 13 , wherein combining the outputs of the plurality of determination models for the image of the target road using the determined weights comprises respectively multiplying outputs of the plurality of determination models by the determined weights.
16. The non-transitory computer-readable recording medium according to claim 15 , wherein the method further comprises:
learning a parameter for determining the weights using training data for learning the weights.
17. The non-transitory computer-readable recording medium according to claim 16 , wherein learning the parameter for determining the eights using the training data for learning the weights comprises:
learning the parameter for determining the weights in accordance with the external environment.
18. The non-transitory computer-readable recording medium according to claim 13 ,
wherein the plurality of determination models respectively determine the outputs as a state of deterioration of the road and structures related to the road.