How to Reconstruct Target Geometry with RGB images
Shilling818 opened this issue · comments
In your paper, you mainly use silhouettes as ground truths. RGB images are used to improve the reconstruction performance. However, when I used your coding and fixed the textures, only RGB images cannot infer the geometry.
Below is the result of my inversion with 24 viewing angles composed of 3 different incident angles and 8 different azimuth angles. What's the problem?
（a） reconstruct with silhouettes
(b) reconstruct with RGB images