Single-view three-dimensional reconstruction method based on cyclic diffusion model
The invention discloses a single-view three-dimensional reconstruction method based on a cyclic diffusion model, and the method comprises the following steps: carrying out the extraction and fusion to obtain a fusion feature vector, and predicting the mean value and variance of an image condition fe...
Saved in:
Main Authors | , , , , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
08.03.2024
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention discloses a single-view three-dimensional reconstruction method based on a cyclic diffusion model, and the method comprises the following steps: carrying out the extraction and fusion to obtain a fusion feature vector, and predicting the mean value and variance of an image condition feature vector; sampling the mean value and variance of the predicted image condition feature vectors to obtain the image condition feature vectors; carrying out farthest point sampling to obtain a three-dimensional model point cloud, and training a denoising network in the diffusion model; the guiding capability of the image condition feature vector to the denoising network is improved through cyclic denoising; inputting a single view to guide a de-noising network to de-noise the pure noise point cloud obeying standard Gaussian distribution step by step, and finally obtaining a three-dimensional model point cloud consistent with the geometric structure of the single view; the single-view three-dimensional reconstruc |
---|---|
Bibliography: | Application Number: CN202311650864 |