Object determining device and program thereof

1. An object determining device comprising at least one processor which implements:

an imager for obtaining an image of the object;

a likelihood value calculator for calculating a first likelihood value for the object shown in the image obtained by the imager by use of the image obtained by the imager and a machine learning system trained so as to respond to a sample image of a sample object that is wearing a covering, the sample image of the sample object that is wearing the covering is applied with a first teacher signal 1 as a first excitatory information and the sample image of the sample object that is not wearing the covering is applied with a first teacher signal 0 as a first inhibitor information, and for calculating a second likelihood value for the object shown in the image obtained by the imager by use of the image obtained by the imager and a machine learning system trained so as to respond to a sample image of a sample object that is not wearing a covering, the sample image of the sample object that is not wearing the covering is applied with a second teacher signal 1 as a second excitatory information and the sample image of the sample object that is wearing the covering is applied with a second teacher signal 0 as a second inhibitor information, the first likelihood value indicating a level of likelihood that the object is wearing the covering and the second likelihood value indicating a level of likelihood that the object is not wearing the covering; and

a determiner for determining whether or not the object, shown in the image obtained by the imager, is wearing a covering, on the basis of a ratio between the first likelihood value and the second likelihood value calculated by the likelihood value calculator.

2. The object determining device according to claim 1 wherein the at least one processor further implements an extractor for extracting, when assuming that the object is wearing the covering, an image within a predetermined area where a covering is estimated to exist, from the image obtained by the imager, and

wherein the likelihood value calculator calculates a first likelihood value for the object shown in the image obtained by the imager by use of the image within the predetermined area extracted by the extractor and a machine learning system trained so as to respond to an image within a predetermined area in a sample image of a sample object that is wearing a covering, the image within the predetermined area in the sample image of the sample object that is wearing the covering is applied with the first teacher signal 1 as the first excitatory information and the image within the predetermined area in the sample image of the sample object that is not wearing the covering is applied with the first teacher signal 0 as the first inhibitor information, and calculates a second likelihood value for the object shown in the image obtained by the imager by use of the image within the predetermined area extracted by the extractor and a machine learning system trained so as to respond to an image within a predetermined area in a sample image of a sample object that is not wearing a covering, the image within the predetermined area in the sample image of the sample object that is not wearing the covering is applied with the second teacher signal 1 as the second excitatory information and the image within the predetermined area in the sample image of the sample object that is wearing the covering is applied with the second teacher signal 0 as the second inhibitor information, and the first likelihood value indicates the level of likelihood that the object is wearing the covering and the second likelihood value indicates the level of likelihood that the object is not wearing the covering.

3. The object determining device according to claim 2 , wherein the determiner determines that the object in the image obtained by the imager is wearing the covering when the ratio of the first likelihood value relative to the second likelihood value exceeds a first threshold, and also determines that the object in the image obtained by the imager is not wearing the covering when the ratio of the first likelihood value relative to the second likelihood value is less than a second threshold, which is set to be smaller than the first threshold.

4. The object determining device according to claim 1 , wherein the determiner determines that the object in the image obtained by the imager is wearing the covering when the ratio of the first likelihood value relative to the second likelihood value exceeds a first threshold, and also determines that the object in the image obtained by the imager is not wearing the covering when the ratio of the first likelihood value relative to the second likelihood value is less than a second threshold, which is set to be smaller than the first threshold.

5. The object determining device according to claim 1 , wherein the object includes a face of a person; and the covering includes one of a mask and colored glasses.

6. A program stored in a non-transitory computer readable medium for a computer executing steps of:

calculating a first likelihood value for an object shown in an image obtained by an imager by use of the image obtained by the imager and a machine learning system trained so as to respond to a sample image of a sample object that is wearing a covering, the sample image of the sample object that is wearing the covering is applied with a first teacher signal 1 as a first excitatory information and the sample image of the sample object that is not wearing the covering is applied with a first teacher signal 0 as a first inhibitor information;

calculating a second likelihood value for the object shown in the image obtained by the imager by use of the image obtained by the imager and a machine learning system trained so as to respond to a sample image of a sample object that is not wearing a covering, the sample image of the sample object that is not wearing the covering is applied with a second teacher signal 1 as a second excitatory information and the sample image of the sample object that is wearing the covering is applied with a second teacher signal 0 as a second inhibitor information, the first likelihood value indicating a level of likelihood that the object is wearing the covering and the second likelihood value indicating a level of likelihood that the object is not wearing the covering; and

determining whether or not the object, shown in the image obtained by the imager, is wearing a covering, on the basis of a ratio between the first likelihood value and the second likelihood value calculated by a likelihood value calculator.

7. The program for the computer according to claim 6 further executing steps of:

extracting, when assuming that the object is wearing the covering, an image within a predetermined area where a covering is estimated to exist, from the image obtained by the imager;

calculating a first likelihood value for the object shown in the image obtained by the imager by use of the image within the predetermined area extracted by an extractor and a machine learning system trained so as to respond to an image within a predetermined area in a sample image of a sample object that is wearing a covering, the image within the predetermined area in the sample image of the sample object that is wearing the covering is applied with the first teacher signal 1 as the first excitatory information and the image within the predetermined area in the sample image of the sample object that is not wearing the covering is applied with the first teacher signal 0 as the first inhibitor information; and

calculating a second likelihood value for the object shown in the image obtained by the imager by use of the image within the predetermined area extracted by the extractor and a machine learning system trained so as to respond to an image within a predetermined area in a sample image of a sample object that is not wearing a covering, the image within the predetermined area in the sample image of the sample object that is not wearing the covering is applied with the second teacher signal 1 as the second excitatory information and the image within the predetermined area in the sample image of the sample object that is wearing the covering is applied with the second teacher signal 0 as the second inhibitor information, the first likelihood value indicating the level of likelihood that the object is wearing the covering and the second likelihood value indicating the level of likelihood that the object is not wearing the covering.

8. The program for the computer according to claim 6 , wherein the object includes a face of a person, and the covering includes one of a mask and colored glasses.

9. The program for the computer according to claim 6 , wherein it is determined that the object in the image obtained by the imager is wearing the covering when the ratio of the first likelihood value relative to the second likelihood value exceeds a first threshold, and also determined that the object in the image obtained by the imager is not wearing the covering when the ratio of the first likelihood value relative to the second likelihood value is less than a second threshold, which is set to be smaller than the first threshold.

10. A method for determining an object, comprising:

a likelihood value calculating process for calculating a first likelihood value for an object shown in an image obtained by an imager for obtaining the object, by use of the image obtained by the imager and a machine learning system trained so as to respond to a sample image of a sample object that is wearing a covering, the sample image of the sample object that is wearing the covering is applied with a first teacher signal 1 as a first excitatory information and the sample image of the sample object that is not wearing the covering is applied with a first teacher signal 0 as a first inhibitor information, and for calculating a second likelihood value for the object shown in the image obtained in by the imager by use of the image obtained in by the imager and a machine learning system trained so as to respond to a sample image of a sample object that is not wearing a covering, the sample image of the sample object that is not wearing the covering is applied with a second teacher signal 1 as a second excitatory information and the sample image of the sample object that is wearing the covering is applied with a second teacher signal 0 as a second inhibitor information, the first likelihood value indicating a level of likelihood that the object is wearing the covering and the second likelihood value indicating a level of likelihood that the object is not wearing the covering; and

a determining process for determining whether or not the object, shown in the image obtained in by the imager, is wearing a covering, on the basis of a ratio between the first likelihood value and the second likelihood value calculated in the likelihood value calculating process.

11. The method for determining the object according to claim 10 , wherein

the image obtained by the imager corresponds to an image within a predetermined area where a covering is estimated to exist, which is extracted when assuming that the object is wearing the covering, and

the first likelihood value is calculated for the object shown in the image obtained by the imager by use of the image within the area extracted and a machine learning system trained so as to respond to an image within the predetermined area in a sample image of a sample object that is wearing a covering, the image within the predetermined area in the sample image of the sample object that is wearing the covering is applied with a first teacher signal 1 as the first excitatory information and the image within the predetermined area in the sample image of the sample object that is not wearing the covering is applied with a first teacher signal 0 as the first inhibitor information, and the second likelihood value is calculated for the object shown in the image obtained by the imager by use of the image within the area extracted and a machine learning system trained so as to respond to an image within the predetermined area in a sample image of a sample object that is not wearing a covering, the image within the predetermined area in the sample image of the sample object that is not wearing the covering is applied with the second teacher signal 1 as the second excitatory information and the image within the predetermined area in the sample image of the sample object that is wearing the covering is applied with the second teacher signal 0 as the second inhibitor information, the first likelihood value indicating the level of likelihood that the object is wearing the covering and the second likelihood value indicating the level of likelihood that the object is not wearing the covering.