I found that the first picture generated was not the image prompt. I feel that it is a problem to add the mse loss of referring image for further image-to-3D generation.
this work adopt canonical camera to generate 3D similiar to the image-prompt, we tried to make the image at the front view, results are worse and more blurry.
I found that the first picture generated was not the image prompt. I feel that it is a problem to add the mse loss of referring image for further image-to-3D generation.