Super-large city streetscape image data sensitive information removal method based on deep learning

The invention relates to an ultra-large city street view image data sensitive information removal method based on deep learning. The method comprises the steps of 1, obtaining a face recognition model; 2, acquiring a license plate recognition model; step 3, extracting an effective data area from the actual streetscape image; 4, face and license plate recognition is carried out on the effective data area in a sliding block mode, and a recognition result is obtained and stored in a recognition json file; step 5, reading the actual street scene image and the corresponding identification json file, and performing Gaussian blur processing on the area of the identification result corresponding to the actual street scene image; and step 6, carrying out accuracy evaluation on the identification result, if the overall accuracy obtained by evaluation is lower than a threshold value, repeating the steps S1 and S2, and continuing to train the model, otherwise, returning to the step S3 to continue to remove the sensitive