Determination device and determining method thereof
A determination device that determines quality of target portion based on sensor data obtained by a sensor measuring the target object, includes one or more processors configured to acquire sensor data representing the target portion, acquire information indicating a changed portion, determine whether the target portion includes the changed portion based on acquired information, determine a first label of the target portion represented in the sensor data by using a determination model learned from a training dataset based on training target portions, the first label representing target portion as one of good, defect, and a defect candidate, accept a second label of the target portion input via a user interface when the target portion includes the changed portion or when the first label of the target portion is determined as the defect candidate, and perform quality determination of the target portion based on the first label and/or the second label.
1. A determination device that determines quality of a target portion of a target object based on sensor data obtained by a sensor measuring the target object, comprising one or more processors configured to:
acquire, from the sensor or a memory, the sensor data representing the target portion;
acquire information indicating a changed portion on the target object;
determine whether the target portion includes the changed portion based on the acquired information;
determine a first label of the target portion represented in the sensor data by using a determination model learned from a training dataset based on training target portions, the first label representing the target portion as one of good, defect, and a defect candidate;
accept a second label of the target portion input via a user interface when the target portion includes the changed portion or when the first label of the target portion is determined as the defect candidate; and
perform quality determination of the target portion based on the first label and/or the second label,
wherein the changed portion is a portion of the target object where a design or a process applied to form the target portion has been changed from that of the training target portions used to form the training dataset.
2. A determination device according to claim 1 , wherein the one or more processors are further configured to:
skip determining the first label of the target portion when the target portion includes the changed portion.
3. A determination device according to claim 1 , wherein the one or more processors are further configured to:
when determining the first label of the target portion that includes the changed portion:
set a threshold for determining the first label between the good and the defect to be stricter than that for determining the first label of the target portion that does not includes the changed portion.
4. A determination device according to claim 1 , wherein the one or more processors are further configured to:
determine whether performing the quality determination of the target portion is necessary based on the sensor data and template data corresponding to the target portion; and
perform the quality determination of the target portion when the quality determination of the target portion is determined to be necessary.
5. A method that determines, when performed by one or more processors, quality of a target portion of a target object based on sensor data obtained by a sensor measuring the target object, comprising:
acquiring, from the sensor or a memory, the sensor data representing the target portion;
acquiring information indicating a changed portion on the target object;
determining whether the target portion includes the changed portion based on the acquired information;
determining a first label of the target portion represented in the sensor data by using a determination model learned from a training dataset based on training target portions, the first label representing the target portion as one of good, defect, and a defect candidate;
accepting a second label of the target portion input via a user interface when the target portion includes the changed portion or when the first label of the target portion is determined as the defect candidate; and
performing quality determination of the target portion based on the first label and/or the second label,
wherein the changed portion is a portion of the target object where a design or a process applied to form the target portion has been changed from that of the training target portions used to form the training dataset.
6. A non-transitory computer readable medium that stores instructions that cause one or more processors to perform a method for determining quality of a target portion of a target object based on sensor data obtained by a sensor measuring the target object, the method comprising:
acquiring, from the sensor or a memory, the sensor data representing the target portion;
acquiring information indicating a changed portion on the target object;
determining whether the target portion includes the changed portion based on the acquired information;
determining a first label of the target portion represented in the sensor data by using a determination model learned from a training dataset based on training target portions, the first label representing the target portion as one of good, defect, and a defect candidate;
accepting a second label of the target portion input via a user interface when the target portion includes the changed portion or when the first label of the target portion is determined as the defect candidate; and
performing quality determination of the target portion based on the first label and/or the second label,
wherein the changed portion is a portion of the target object where a design or a process applied to form the target portion has been changed from that of the training target portions used to form the training dataset.