Automated optical inspection and classification apparatus based on a deep learning system and training apparatus thereof

An automated optical inspection and classification apparatus based on a deep learning system, to be used to detect a defect of an object through an image of the object, the automated optical inspection and classification apparatus comprising:

• a processor for executing the deep learning system after loading data from a storage unit, wherein the deep learning system includes:
• an input layer for receiving and normalizing the image of the object into an input image;
• a neural network layer group for extracting the input image and thereby obtaining a plurality of image features, wherein the neural network layer group includes at least one neural network sub-block layer and a max-pooling layer;
• and
• a fully connected-layer group with at least one fully connected layer for performing weight-based classification and outputting an inspection result;
• wherein the neural network sub-block layer includes:
• a first operational subset and a second operational subset that share a common input,
• an element-wise (eltwise) layer for performing a residual-related operation on an output of the first operational subset and an output of the second operational subset, and
• a rectified linear unit (ReLU) for performing a nonlinear operation on an output of the eltwise layer.