Machine learning device, inspection device and machine learning method

1. A machine learning device for creating training data to be used in machine learning, the machine learning device comprising:

an image input unit that inputs an image which was obtained by capturing an inspection target on which a symbol indicating a defect may be marked; and

a creation unit that creates the training data based on the inputted image,

wherein the creation unit:

creates training data consisting of a training image which is the image as inputted, and a label that retains a value of OK which signifies not having a defect, in a case of there not being the symbol in the image inputted; and

creates training data consisting of a training image generated based on the image inputted, and a label that retains a value of NG signifying having a defect, in a case of there being the symbol in the image inputted.

2. The machine learning device according to claim 1 ,

wherein the creation unit creates training data consisting of a training image generated based on the image inputted, a position of a defect indicated by the symbol, and a label retaining the value of NG, in a case of there being the symbol in the image inputted.

3. The machine learning device according to claim 1 , wherein the symbol indicating the defect is a different symbol depending on a type of the defect, and

wherein the creation unit creates training data consisting of a training image generated based on the image inputted, a type of defect indicated by the symbol, and a label retaining the value of NG, in a case of there being the symbol in the image inputted.

4. The machine learning device according to claim 3 , wherein a scratch, stain and chip of the inspection target are included as types of the defect.

5. The machine learning device according to claim 3 , wherein the different symbol is a symbol having a different reflective optical wavelength from another symbol.

6. The machine learning device according to claim 1 , wherein the creation unit:

generates a training image by removing the symbol from an image generated based on the image inputted, in a case of there being the symbol in the image inputted, and

creates training data consisting of the training image thus generated, and a label retaining the value of NG.

7. The machine learning device according to claim 6 ,

wherein the creation unit generates a training image by cutting out only inside of a region indicated by the symbol from the image inputted, in a case of there being the symbol in the image inputted.

8. The machine learning device according to claim 6 ,

wherein the creation unit generates a training image by replacing pixels constituting the symbol from the image inputted with peripheral pixels of the symbol, in a case of there being the symbol in the image inputted.

9. The machine learning device according to claim 1 ,

wherein the creation unit creates training data consisting of a training image generated based on the image inputted, a label retaining a value of NG signifying having a defect, and a mask image that masks the symbol, in a case of there being the symbol in the image inputted.

10. The machine learning device according to claim 9 , wherein the mask image is an image setting pixels constituting the symbol as 1, and setting pixels other than the symbol as 0.

11. The machine learning device according to claim 1 , wherein the image input unit inputs a marking image obtained by capturing an inspection target on which a symbol indicating a defect is marked in a state illuminating with light of a wavelength whereby the symbol is visible, and a training image obtained by capturing in a state illuminating with light of a wavelength whereby the symbol is not visible, and

wherein the creation unit creates training data consisting of information of a defect acquired from the marking image, and the training image.

12. An inspection device comprising:

the machine learning device according to claim 1 ; and

a storage unit that stores training data created by the machine learning device,

wherein the machine learning device includes a learning unit which learned from training data stored by the storage unit, and outputs a determination result of a defect based on an image of the inspection target inputted by the image input unit.

13. An inspection device comprising:

the machine learning device according to claim 11 ; and

a storage unit that stores training data created by the machine learning device,

wherein the image input unit inputs an image obtained by capturing in a state illuminating with light of a wavelength whereby the symbol indicating a defect is not visible, and

wherein the machine learning device includes a learning unit which learned from training data stored by the storage unit, and outputs a determination result of a defect based on an image of the inspection target inputted by the image input unit.

14. A machine learning method for creating training data to be used in machine learning, the method comprising the steps of:

inspecting an inspection target, and when a defect is found, marking a symbol on a defect location;

inputting an image obtained by capturing the inspection target; and

creating the training data based on the image inputted,

wherein the step of creating the training data includes:

a step of creating training data consisting of a training image which is the image as inputted, and a label that retains a value of OK signifying not having a defect, in a case of there not being the symbol in the image inputted; and

a step of creating training data consisting of a training image generated based on the image inputted, and a label that retains a value of NG signifying having a defect, in a case of there being the symbol in the image inputted.