Deep learning-based overlay key centering system and method thereof

A method of centering a deep learning-based overlay comprising:

• a step of collecting an input data set from at least one device, the input data set comprising measurement image data of an overlay key and label data including information on a position and bounding box size of the overlay;
• and
• a step of training the model by inputting the input data set to a model for deep learning,
• wherein the label data comprises at least one of a type of the overlay key to be trained, a value obtained by dividing a x-coordinate value among center coordinates (x, y) of the overlay key to width value of measured image of the overlay key, a value obtained by dividing a y-coordinate value among center coordinates (x, y) of the overlay key to height value of measured image of the overlay key, a value obtained by dividing width of the bounding box of the overlay key to width value of measured image of the overlay key, and a value obtained by dividing height of the bounding box of the overlay key to height value of measured image of the overlay key, and
• wherein the step of training the model comprises,
• a step of calculating a loss function by comparing result data predicted by the model with label data;
• and
• a step of optimizing an algorithm of the model by modifying a weight of the model so that a loss value calculated with the loss function may become smaller than a reference value.