Anti-fact sample generation method, model adjustment method, equipment and medium

The embodiment of the invention discloses an anti-fact sample generation method, a model adjustment method, equipment and a medium. The method comprises the following steps: constructing a mirror image network model symmetrical to an original machine learning network model according to the original machine learning network model, and generating an anti-fact network model; performing a model training test on the original machine learning network model through the test sample set, and obtaining a test result; and taking a target sample which is wrongly identified in the test result as the input of the anti-fact network model to obtain an anti-fact sample output by the anti-fact network model. The test result of the test stage can be automatically fed back to the learning stage, the anti-fact sample can be generated to expand the learning sample, the problem of small samples is solved, then model deviation correction can be automatically achieved under the condition of not depending on manual participation, the