Model input size determination method, electronic device and storage medium

An electronic device comprising a memory and a processor, the memory stores at least one computer-readable instruction, which when executed by the processor causes the processor to:

• acquire a test image set, the test image set includes a plurality of test images and a defect result corresponding to each test image;
• encode each test image to obtain an encoding vector;
• decode the encoding vector to obtain a reconstructed image corresponding to the test image and compare the reconstructed image with the corresponding test image to obtain a reconstruction error;
• segment the encoding vector according to a plurality of preset sizes to obtain a plurality of sub-vectors corresponding to each preset size;
• input the plurality of sub-vectors corresponding to each preset size into a Gaussian mixture model to obtain a plurality of sub-probabilities, and determine an estimated probability according to the plurality of sub-probabilities corresponding to each preset size;
• determine a test error according to the estimated probability and the reconstruction error corresponding to each preset size, and obtain a detection result in the test image corresponding to each preset size according to the test error and the corresponding error threshold;
• determine an accuracy according to the detection result and the defect result corresponding to each preset size and select an input size from the plurality of preset sizes according to the accuracy corresponding to each preset size.

A non-transitory storage medium having stored thereon at least one computer-readable instructions that, when the at least one computer-readable instructions are executed by a processor to implement a model input size determination method:

• acquiring a test image set, the test image set includes a plurality of test images and a defect result corresponding to each test image;
• encoding each test image to obtain an encoding vector;
• decoding the encoding vector to obtain a reconstructed image corresponding to the test image, comparing the reconstructed image with the corresponding test image to obtain a reconstruction error;
• segmenting the encoding vector according to a plurality of preset sizes to obtain a plurality of sub-vectors corresponding to each preset size;
• inputting the plurality of sub-vectors corresponding to each preset size into a Gaussian mixture model to obtain a plurality of sub-probabilities, and determining an estimated probability according to the plurality of sub-probabilities corresponding to each preset size;
• determining a test error according to the estimated probability and the reconstruction error corresponding to each preset size, and obtaining a detection result in the test image corresponding to each preset size according to the test error and the corresponding error threshold;
• determining an accuracy according to the detection result and the defect result corresponding to each preset size, and selecting an input size from the plurality of preset sizes according to the accuracy corresponding to each preset size.