[HUANG XI-XUAN, HUANG ZHE-XUAN, ZHAN HAO-ZHONG, ZHANG SHU-XIU, ZHANG SHUN-BO, LIN JUN-YOU] TW An image inspection system is provided. The image inspection system includes a transceiver, a storage and a processor. The transceiver is used to receive a test image. The storage is connected to the transceiver, used to store the test image. The processor is connected to the storage, and the processor includes a setting circuit and an operation circuit. The setting circuit is used to provide an operation interface so that the user can self-define a plurality of areas on the test image. The operation circuit is connected to the setting circuit, used to use different machine learning model for each of the areas to detect.

[HUANG XI-XUAN, HUANG ZHE-XUAN, ZHAN HAO-ZHONG, ZHANG SHU-XIU, ZHANG SHUN-BO, LIN JUN-YOU] TW A character recognition system is provided. The character recognition system includes a database and a processor. The database stores a large amount of data that has completed character classification and marking as the first training data. The processor is connected to the database and includes a first layer training circuit, a correction circuit, and a second layer training circuit. The first layer training circuit is used to make the machine learning model use the first training data for training. The correction circuit is connected to the first layer training circuit, used to manually correct at least one character which is incorrectly recognized by the machine learning model, and generate at least one mark in the test data to indicate what the corresponding at least one character is. The second layer training circuit is connected to the correction circuit, used to make the machine learning model further use the test data with the at least one mark as the second training data for training.

[HUANG ZHE-XUAN, HUANG XI-XUAN, ZHANG SHU-XIU, ZHAN HAO-ZHONG] TW

[HUANG XI-XUAN, HUANG ZHE-XUAN, ZHANG SHU-XIU, ZHAN HAO-ZHONG, ZHANG SHUN-BO, ZHOU JIA-ROU] TW