search

facebookresearch / pifuhd

High-Resolution 3D Human Digitization from A Single Image.
Other
9.5k stars 1.44k forks source link

Including original image as texture #99

Closed stevesmit closed 3 years ago

stevesmit commented 3 years ago

In the paper, I note in figure 2 that the original image appears to have been placed on the mesh. Is this possible with the code in this repo? If so, could you point me in the right direction?

shunsukesaito commented 3 years ago

The resulting meshes are in the pixel coordinates after cropping. So if you render the meshes with orthogonal projection, you can overlay the rendering on the input image as in Fig. 2. Currently no plan to release this rendering code itself in the repo.

merlinyx commented 1 year ago

Hi! I've been trying to do the orthogonal projection based on the camera paramters in camera.py, or through reverting the reconstruction process to get back to the pixel coordinates but the resulting projection does not align with the input image (meaning if I project the pifuhd output onto the input image coordinate space, the points do not fall on the human in the image and the scale is a bit off). Do you know what might be the issue? Was there any hidden or default transformation to make them align? Thank you!