Object detection device, vehicle control system, object detection method, and non-transitory computer readable medium

1. An object detection device comprising:

a processor, said processor comprising:

a position region detecting part configured to use a first neural network to detect a position region of an object in the image,

a large attribute identification part configured to use a second neural network to identify a large attribute of the object,

a small attribute identification part configured to use a third neural network to identify a small attribute of the object which is a lower concept of the large attribute, and

an object judging part configured to judge a result of detection of the object,

wherein

the object judging part is configured to judge that a result of identification of the small attribute is the result of detection if a confidence of the result of identification of the small attribute by the small attribute identification part is equal to or more than a threshold value, and judge the result of detection based on a result of identification of the large attribute if the confidence is less than the threshold value,

wherein the object judging part is configured to judge that a speed limit sign of a slowest speed in candidates of the small attribute is the result of detection if the confidence is less than the threshold value and the result of identification of the large attribute by the large attribute identification part is a speed limit sign.

2. The object detection device according to claim 1 , wherein the object judging part is configured to judge the result of detection based on the result of identification of the large attribute and a distribution of confidence of the small attribute if the confidence is less than the threshold value.

3. A vehicle control system comprising

an object detection device according to claim 1 ,

a drive planning part configured to create a drive plan of a vehicle based on the result of detection of the object, and

a vehicle control part configured to control the vehicle so that the vehicle drives in accordance with a drive plan prepared by the drive planning part.

4. A method of detection of an object comprising:

providing a processor having a first neural network, a second neural network and a third neural network,

using the first neural network to detect a position region of an object in an image,

using the second neural network to identify a large attribute of the object,

using the third neural network to identify a small attribute of the object which is a lower concept of the large attribute,

when a confidence of a result of identification of the small attribute is equal to or more than a threshold value, judging that the result of identification of the small attribute is a result of detection of the object and, when the confidence is less than the threshold value, judging the result of detection based on a result of identification of the large attribute, and judging that a speed limit sign of a slowest speed in candidates of the small attribute is the result of detection if the confidence is less than the threshold value and the result of identification of the large attribute by the large attribute identification part is a speed limit sign.

5. A non-transitory computer-readable medium storing an object detection use computer program making a computer having a processor:

use a first neural network to detect a position region of an object in an image,

use a second neural network to identify a large attribute of the object,

use a third neural network to identify a small attribute of the object which is a lower concept of the large attribute, and,

when a confidence of a result of identification of the small attribute is equal to or more than a threshold value, judge that the result of identification of the small attribute is a result of detection of the object and, when the confidence is less than the threshold value, judge the result of detection based on a result of identification of the large attribute, and judge that a speed limit sign of a slowest speed in candidates of the small attribute is the result of detection if the confidence is less than the threshold value and the result of identification of the large attribute by the large attribute identification part is a speed limit sign.

6. An object detection device having a processor configured to:

use a first neural network to detect a position region of an object in the image,

use a second neural network to identify a large attribute of the object,

use a third neural network to identify a small attribute of the object which is a lower concept of the large attribute,

judge a result of detection of the object, judge that a result of identification of the small attribute is the result of detection if a confidence of the result of identification of the small attribute is equal to or more than a threshold value, and judge the result of detection based on a result of identification of the large attribute if the confidence is less than the threshold value, and

judge that a speed limit sign of a slowest speed in candidates of the small attribute is the result of detection if the confidence is less than the threshold value and the result of identification of the large attribute by the large attribute identification part is a speed limit sign.