I tried to easily create a high-precision 3D image with one photo [3]. MiDaS of feat. Intel-isl only depth without permission.
First output
** Arbitrarily depth only feat. Intel-isl. The result of **.
Here, the following two github outputs are mixed. For the former, only the calculation of depth is replaced with the latter. There is no deep meaning. I expected to get good results, For the time being, ** not much good has happened. ** **
| Division of roles | github |
|---|---|
| Other than depth | https://github.com/vt-vl-lab/3d-photo-inpainting You can create a 3D image from one photo |
| depth | https://github.com/intel-isl/MiDaS Depth can be calculated from one photo |
Overview
** One photo ** ** High-precision 3D image ↓ Easy creation **.
I tried to easily create a high-precision 3D image with one photo [1]. (Depth can now be edited in PNG.) See the article.
here,
| The depth can be changed instead of using the original processing. intel-Created using isl's github (PNG)And tried to adopt it (↑ This is all the procedure) |
|---|
⇒ ** Collaboration without permission ** ⇒⇒ ** I wonder if I can do it **
Below is the original paper and github.
https://arxiv.org/pdf/2004.04727.pdf Paper "3D Photography using Context-aware Layered Depth Inpainting" Meng-Li Shih1 and others
github is below. https://github.com/vt-vl-lab/3d-photo-inpainting
** Used to create depth information (PNG) ** The github of intel-isl is below. https://github.com/intel-isl/MiDaS
result
The first thing shown above is the ** result **. The original (= not replaced with MiDaS of intel-isl) is I tried to easily create a high-precision 3D image with one photo [1]. (Depth can now be edited in PNG.) See the article.
Is the condition a little off? ** Not good ** ** However, there are some areas that have improved. ** I feel.
from now on
I would like to consider it a little more carefully. Simple, I don't know the following two.
- Absolute value of PNG. (It was good because it was a gray scale that somehow understood the atmosphere.)
- I wonder if it's easy to see the vertical plane (wall) or horizontal plane. .. .. (Why can't you recognize it well?)
If you have any comments, please let us know.
reference. I tried to easily create a high-precision 3D image with one photo [2]. (Try processing depth with numpy) It is described above, but just in case ↓ I tried to easily create a high-precision 3D image with one photo [1]. (Depth can now be edited in PNG.)
Recommended Posts