Obstacle removal and three-dimensional reconstruction method based on deep learning

The embodiment of the invention discloses a shielding object removal and three-dimensional reconstruction method based on deep learning. A specific embodiment of the method comprises the steps of performing recognition processing on a two-dimensional shielded face image to obtain a mask grey-scale image and a mask contour image corresponding to the two-dimensional shielded face image; inputting the two-dimensional shielded face image into a pre-trained mask image generation network to generate a mask image corresponding to the two-dimensional shielded face image; inputting the mask grey-scale image, the mask contour image and the mask image into a pre-trained contour generator to generate a mask-free complete contour composite image; inputting the two-dimensional sheltered face image and the maskless complete contour composite image into a pre-trained face rendering network to generate a two-dimensional unsheltered face rendering image; and carrying out conversion processing on the two-dimensional non-occlusi