Abstract: The present invention provides an automated optical inspection and classification apparatus based on a deep learning system, comprising a camera and a processor. The processor executes a deep learning system after loading data from a storage unit and the processor, and comprises an input layer, a neural network layer group, and a fully connected-layer group. The neural network layer group is for extracting to an input image and thereby obtaining a plurality of image features. The fully connected-layer group is for performing weight-based classification and outputting an inspection result.

Abstract: An optical inspection method for an optical inspection device comprising an optical lens is provided according to an embodiment of the disclosure. The optical inspection method includes: obtaining a first image of an object by the optical lens; performing an edge detection on the first image to obtain a second image comprising an edge pattern; and performing a defect inspection operation on the second image based on a neural network architecture to inspect a defect pattern in the second image. In addition, an optical inspection device and an optical inspection system are provided according to embodiments of the disclosure.

Abstract: A defect inspection system, connected to an automatic visual inspection device, is provided, including the followings. A re-inspection server (VRS) receives a defect image and a defect location. A training terminal stores trained modules. A classification terminal receives the defect image and the defect location, reads a target trained module corresponding to the defect image, classifies the defect image according to the target trained module to obtain a labeled defect image, and sends the labeled defect image to the VRS. A re-inspection terminal receives the labeled defect image from the VRS, and sends a verified operation corresponding to the labeled defect image to the VRS. A labeling re-inspection terminal receives the verified operation and the labeled defect image, and a labeling result corresponding to the labeled defect image. The VRS sends the labeling result and the labeled defect image to the training terminal to train a corresponding training module.

Abstract: The present invention is an automated optical inspection method using deep learning, comprising the steps of: providing a plurality of paired image combinations, wherein each said paired image combination includes at least one defect-free image and at least one defect-containing image corresponding to the defect-free image; providing a convolutional neural network to start a training mode of the convolutional neural network; inputting the plurality of paired image combinations into the convolutional neural network, and adjusting a weight of at least one fully connected layer of the convolutional neural network through backpropagation to complete the training mode of the convolutional neural network; and performing an optical inspection process using the trained convolutional neural network.

Abstract: A method for evaluating an efficiency of a manual inspection for a defect pattern is provided according to an embodiment of the disclosure, which comprises: enabling an evaluation program; loading a test image automatically by the enabled evaluation program and displaying the test image in a user interface; detecting a user behavior of a user after the user watches the test image; generating original data according to the user behavior, wherein the original data reflects at least one of whether the user identifies the defect pattern in the test image and a type of the defect pattern identified by the user; and performing a quantitative operation on the original data to generate evaluation data corresponding to the efficiency of the manual inspection, wherein the evaluation data reflects an evaluation result corresponding to the efficiency of the manual inspection.

Abstract: An optical inspection method for an optical inspection device comprising an optical lens is provided according to an embodiment of the disclosure. The optical inspection method includes: obtaining a first image of an object by the optical lens; performing an edge detection on the first image to obtain a second image comprising an edge pattern; and performing a defect inspection operation on the second image based on a neural network architecture to inspect a defect pattern in the second image. In addition, an optical inspection device and an optical inspection system are provided according to embodiments of the disclosure.

Abstract: A training method of an iterative deep learning system, comprising the steps of: providing at least one unlabeled image of an object; labeling the unlabeled image of the object to produce a labeled image of the object; storing the labeled image of the object into an image database if the deep learning system is being trained with a labeled version of the unlabeled image of the object for a first time; identifying the labeled image of the object in the image database and outputting an identification result through the deep learning system; and training the deep learning system according to an error between output and expected output of the identification result; and providing another unlabeled image of the object to the trained deep learning system to identify said another unlabeled image of the object and to produce an inspection result; wherein determining whether the inspection result has satisfied a qualification condition and terminating training if the inspection result has satisfied the qualification co

Abstract: A labeling system for defect classification having a storage unit and a processing unit is provided. The storage unit stores a defect classification module, a defect labeling module, and a catalog generation module. The processing unit performs the modules aforementioned. The defect classification module is configured to provide a defect type information. The defect labeling module marks a defect type label to at least one test object image according to an image feature and the defect type information on the at least one test object image. The catalog generation module receives the labeled test object images and moves test object images having the same defect type label to a corresponding file catalog.

Abstract: The present invention provides an automated optical inspection and classification apparatus based on a deep learning system, comprising a camera and a processor. The processor executes a deep learning system after loading data from a storage unit and the processor, and comprises an input layer, a neural network layer group, and a fully connected-layer group. The neural network layer group is for extracting to an input image and thereby obtaining a plurality of image features. The fully connected-layer group is for performing weight-based classification and outputting an inspection result.

Abstract: A defect inspection system, connected to an automatic visual inspection device, is provided, including the followings. A re-inspection server (VRS) receives a defect image and a defect location. A training terminal stores trained modules. A classification terminal receives the defect image and the defect location, reads a target trained module corresponding to the defect image, classifies the defect image according to the target trained module to obtain a labeled defect image, and sends the labeled defect image to the VRS. A re-inspection terminal receives the labeled defect image from the VRS, and sends a verified operation corresponding to the labeled defect image to the VRS. A labeling re-inspection terminal receives the verified operation and the labeled defect image, and a labeling result corresponding to the labeled defect image. The VRS sends the labeling result and the labeled defect image to the training terminal to train a corresponding training module.

Abstract: A method for evaluating an efficiency of a manual inspection for a defect pattern is provided according to an embodiment of the disclosure, which comprises: enabling an evaluation program; loading a test image automatically by the enabled evaluation program and displaying the test image in a user interface; detecting a user behavior of a user after the user watches the test image; generating original data according to the user behavior, wherein the original data reflects at least one of whether the user identifies the defect pattern in the test image and a type of the defect pattern identified by the user; and performing a quantitative operation on the original data to generate evaluation data corresponding to the efficiency of the manual inspection, wherein the evaluation data reflects an evaluation result corresponding to the efficiency of the manual inspection.

Abstract: The present invention is an automated optical inspection method using deep learning, comprising the steps of: providing a plurality of paired image combinations, wherein each said paired image combination includes at least one defect-free image and at least one defect-containing image corresponding to the defect-free image; providing a convolutional neural network to start a training mode of the convolutional neural network; inputting the plurality of paired image combinations into the convolutional neural network, and adjusting a weight of at least one fully connected layer of the convolutional neural network through backpropagation to complete the training mode of the convolutional neural network; and performing an optical inspection process using the trained convolutional neural network.