Image inspection device

Abstract

The present invention provides an image inspection device that can detect minute defects in input images without excessively increasing the resolution of the input images that can be processed by a network of machine learning models. [Solution] The image inspection device (1) includes a display unit (53) that displays a work image in which the workpiece is shown, an information generation unit (54) that generates defect information corresponding to a work image to be used as a training image based on annotations that identify the defect regions included in the displayed work image, a learning execution unit (54) that trains a machine learning model (segmentation model), and an inspection execution unit (13a) that executes the trained machine learning model and displays the defect regions of the inspection image in which the workpiece to be inspected is shown on the display unit as an execution result. The learning execution unit determines whether or not to divide the training image according to the detection sensitivity setting of the defect region, and when it determines to divide the training image, it divides the training image into predetermined batch sizes and inputs them to the machine learning model.

Claims

1. An image inspection device that runs a machine learning model in which parameters are updated by machine learning based on training images provided by the user, A display unit that shows a work image of the workpiece, An information generation unit generates defect information corresponding to the work image which will be the learning image, based on annotations that identify defect regions included in the displayed work image. A learning execution unit that learns the machine learning model, which is a segmentation model that classifies image data pixel by pixel, An inspection execution unit that executes the trained machine learning model and displays the defect region of the inspection image showing the workpiece to be inspected on the display unit as an execution result, Equipped with, The image inspection device includes a learning execution unit which determines whether or not to divide the learning image according to the detection sensitivity setting of the defect region, and when it determines to divide the learning image, divides the learning image into predetermined batch sizes and inputs them into the machine learning model.

2. The image inspection apparatus according to claim 1, wherein the learning execution unit reduces the resolution of the learning image according to the detection sensitivity setting of the defect region, and divides the reduced-resolution learning image into predetermined batch sizes.

3. The image inspection apparatus according to claim 1, wherein the learning execution unit automatically sets the detection sensitivity of the defect region based on the defect information.

4. The image inspection apparatus according to claim 3, wherein the learning execution unit automatically sets the detection sensitivity of the defect region based on the minimum size of the defect region included in the defect information.

5. The image inspection apparatus according to claim 3 or 4, wherein the learning execution unit determines that the learning image will not be divided when automatically setting the detection sensitivity.

6. The image inspection apparatus according to claim 3 or 4, wherein the learning execution unit can change the detection sensitivity of the automatically set defect region by user operation.

7. The image inspection apparatus according to claim 1, wherein the learning execution unit causes the display unit to display a warning message when multiple sizes of the defect region are included in the defect information.

8. The image inspection apparatus according to claim 7, wherein the learning execution unit causes the display unit to display the warning message when the relative ratio between the size of the first defect region and the size of the second defect region is greater than or equal to a certain value.

9. The image inspection apparatus according to claim 1, wherein the learning execution unit divides the learning image into a plurality of divided images while providing overlapping regions between adjacent divided images.

10. The image inspection apparatus according to claim 3 or claim 4, wherein the display unit displays a size indicator, which represents the size of the defect region according to the automatic setting of the detection sensitivity, superimposed on the learning image.

11. The image inspection apparatus according to claim 3 or 4, wherein the learning execution unit can change the detection sensitivity setting of the automatically set defect region by user operation to change the size of the size display.

Images