Analysis of the results of neural networks trained for the detection of defects in plastic plates for packaging. It is about carrying out the supervision and validation of a learning system for defects in plastic sheets intended for packaging. Both algorithms based on artificial intelligence and classical vision algorithms are analyzed.