Information processing device, learning processing method, learning device, and object recognition device
An information processing device, a learning processing method, a learning device and an object recognition device capable of improving learning processing accuracy are provided. An information processing device causes a computer to generate a distortion-corrected equivalent image on the basis of a first captured image including a subject having predetermined distortion generated due to optical operations during imaging, to generate first teacher data in which relevant information about the target object has been added for the equivalent image in which an image region for the target object has been designated, to convert the equivalent image included in the first teacher data into a distorted image having predetermined distortion to generate second teacher data and to generate a learning model which outputs a result obtained by identifying the target object included in the second captured image when the second captured image having predetermined distortion generated therein is input, on the basis of the second teacher data.
1. An information processing device causing a computer to:
generate a distortion-corrected equivalent image on the basis of a first captured image including a subject having predetermined distortion generated due to optical operations during imaging;
generate first teacher data in which relevant information about a target object has been added for the equivalent image in which an image region for the target object has been designated;
convert the equivalent image included in the first teacher data into a distorted image having predetermined distortion to generate second teacher data; and
generate a learning model which outputs a result obtained by identifying the target object included in a second captured image when the second captured image having predetermined distortion generated therein is input, on the basis of the second teacher data.
2. The information processing device according to claim 1 , wherein the first captured image and the second captured image are images captured by an imager including a fisheye lens.
3. The information processing device according to claim 1 , further causing the computer to surround the image region of the target object with a rectangular frame to designate the image region corresponding to the target object.
4. The information processing device according to claim 3 , further causing the computer to convert the equivalent image included in the first teacher data into the distorted image and to convert the shape of the rectangular frame into a shape having distortion according to the position on the equivalent image at which the rectangular frame has been designated when the second teacher data is generated.
5. A learning processing method executed by a computer, comprising:
generating a distortion-corrected equivalent image on the basis of a first captured image including a subject having predetermined distortion generated due to optical operations during imaging;
generating first teacher data in which relevant information about a target object has been added for the equivalent image in which an image region for the target object has been designated;
converting the equivalent image included in the first teacher data into a distorted image having predetermined distortion to generate second teacher data; and
generating a learning model which outputs a result obtained by identifying the target object included in a second captured image when the second captured image having predetermined distortion generated therein is input, on the basis of the second teacher data.
6. A learning device comprising:
an image corrector which generates a distortion-corrected equivalent image on the basis of a first captured image including a subject having predetermined distortion generated due to optical operations during imaging;
a first teacher data generator which generates first teacher data in which relevant information about a target object has been added for the equivalent image in which an image region for the target object has been designated;
a second teacher data generator which converts the equivalent image included in the first teacher data into a distorted image having predetermined distortion to generate second teacher data; and
a learning model generator which generates a learning model which outputs a result obtained by identifying the target object included in a second captured image when the second captured image having the predetermined distortion generated therein is input, on the basis of the second teacher data.
7. An object recognition device which recognizes an object using a learning model generated by a computer which:
generates the distortion-corrected equivalent image on the basis of a first captured image including a subject having predetermined distortion generated due to optical operations during imaging;
generates first teacher data in which relevant information about the target object has been added for the equivalent image in which an image region for the target object has been designated;
converts the equivalent image included in the first teacher data into a distorted image having predetermined distortion to generate second teacher data; and
generates the learning model which outputs a result obtained by identifying the target object included in a second captured image when the second captured image having predetermined distortion generated therein is input, on the basis of the second teacher data.
8. The object recognition device according to claim 7 , further comprising a recognizer which recognizes the target object included in a third captured image captured by an imager in which the predetermined distortion is generated on the basis of the third captured image having distortion that has not been corrected.