Object detection device, learned model generation method, and recording medium
In an object detection device, the plurality of object detection units output a score indicating a probability that a predetermined object exists for each partial region set with respect to inputted image data. The weight computation unit uses weight computation parameters to compute a weight for each of the plurality of object detection units on a basis of the image data and outputs of the plurality of object detection units, the weight being used when the scores outputted by the plurality of object detection units are merged. The merging unit merges the scores outputted by the plurality of object detection units for each partial region according to the weights computed by the weight computation unit. The first loss computation unit computes a difference between a ground truth label of the image data and the score merged by the merging unit as a first loss. Then, the first parameter correction unit corrects the weight computation parameters so as to reduce the first loss.
1. An object detection device comprising:
a memory configured to store instructions; and
one or more processors configured to execute the instructions to:
output a score indicating a probability that a predetermined object exists for each partial region set with respect to inputted image data; data, from a plurality of object detection units;
use weight computation parameters to compute a weight for each of the plurality of object detection units on a basis of the image data and outputs of the plurality of object detection units, the weight being used when the scores outputted by the plurality of object detection units are merged;
merge the scores outputted by the plurality of object detection units for each partial region according to the computed weights;
compute a difference between a ground truth label of the image data and the merged score as a first loss; and
correct the weight computation parameters so as to reduce the first loss.
2. The object detection device according to claim 1 ,
wherein the one or processors compute a single weight with respect to the image data as a whole, and
wherein the one or more processors merge the scores outputted by the plurality of object detection units according to the single weight.
3. The object detection device according to claim 1 ,
wherein the one or processors compute the weight for each partial region of the image data, and
wherein the one or processors merge the scores outputted by the plurality of object detection units according to the weight computed for each partial region.
4. The object detection device according to claim 1 , wherein the one or processors multiply the scores outputted by the plurality of object detection units by the weight computed for each object detection unit, add the multiplied scores together, and calculate an average value.
5. A learned model generation method comprising:
outputting, from a plurality of object detection units, a score indicating a probability that a predetermined object exists for each partial region set with respect to inputted image data;
using weight computation parameters to compute a weight for each of the plurality of object detection units on a basis of the image data and outputs of the plurality of object detection units, the weight being used when the scores outputted by the plurality of object detection units are merged;
merging the scores outputted by the plurality of object detection units for each partial region according to the computed weights;
computing a difference between a ground truth label of the image data and the merged score as a first loss; and
correcting the weight computation parameters so as to reduce the first loss.
6. A recording medium storing a program causing a computer to execute the method according to claim 5 .
7. An object detection device comprising:
a memory configured to store instructions; and
one or more processors configured to execute the instructions to:
output a score indicating probability that a predetermined object exists, for each partial region set to image data inputted, from a plurality of object detection units;
compute weights for merging the scores outputted by the plurality of object detection units, using weight computation parameters, based on the image data and outputs of the plurality of object detection units;
merge the scores outputted by the plurality of object detection units, for each partial region, with the computed weights;
output a score indicating probability that the predetermined object exists, for each partial region set to the image data, from a target model object detection unit;
compute a second loss indicating a difference of the score of the target model object detection unit from a ground truth label of the image data and the merged score; and
correct parameters of the target model object detection unit to reduce the second loss.
8. The object detection device according to claim 7 , the one or more processors are further configured to execute the instructions to:
compute a third loss indicating a difference between the ground truth label and the merged score; and
correct the weight computation parameters to reduce the third loss.
9. The object detection device according to claim 8 , wherein the one or more processors correct the weight computation parameters based on the first second and the third loss.
10. The object detection device according to claim 7 ,
wherein the one or more processors estimate estimates the weights outputted by the weight computation unit based on the image data by the target model object detection unit,
wherein the one or more processors compute a fourth loss indicating a difference between the outputted weights and the estimated weights, and
wherein the one or more processors correct parameters of the target model object detection unit to reduce the second loss and the fourth loss.
11. The learned model generation method according to claim 5 , further comprising:
outputting a score indicating probability that the predetermined object exists, for each partial region set to the image data, from a target model object detection unit;
computing a second loss indicating a difference of the score of the target model object detection unit from a ground truth label of the image data and the merged score; and
correcting parameters of the target model object detection unit to reduce the second loss.
12. A recording medium recording a program causing a computer to execute the method according to claim 5 , and
outputting a score indicating probability that the predetermined object exists, for each partial region set to the image data, from a target model object detection unit;
computing a second loss indicating a difference of the score of the target model object detection unit from a ground truth label of the image data and the merged score; and
correcting parameters of the target model object detection unit to reduce the second loss.