Real-time verification method for deep learning model reasoning

The invention discloses a real-time verification method for deep learning model reasoning, and the method comprises the steps: determining each piece of verification data and result data corresponding to each piece of verification data, then obtaining to-be-detected data related to space edge calculation in response to a prediction instruction of a user, adding the verification data into the to-be-detected data to obtain batch data, and carrying out the calculation of the batch data. And inputting the batch data into the prediction model to obtain a prediction result corresponding to the to-be-detected data and a prediction result corresponding to the at least one piece of verification data. And further, according to a prediction result corresponding to the verification data and result data corresponding to the verification data, determining a verification result, if the verification result is determined to indicate that the verification is passed, returning the prediction result corresponding to the to-be-de