Trying to be constructive but this is merely a base for testing based on very academical work. First:
-
The input images need to be in a very specific format. Almost wireframes. Otherwise it doesn't produce good results. If I have to prepare the wireframe why shouldn't go the full way for the 3D?
-
Not even the examples given in test work, hallucinating and producing incoherent results.
-
Last commits are from 2025.
Worth checking if you are an image processing specialist trying to earn your place at some University... but otherwise totally unusable in production environments.
Trying to be constructive but this is merely a base for testing based on very academical work. First:
The input images need to be in a very specific format. Almost wireframes. Otherwise it doesn't produce good results. If I have to prepare the wireframe why shouldn't go the full way for the 3D?
Not even the examples given in test work, hallucinating and producing incoherent results.
Last commits are from 2025.
Worth checking if you are an image processing specialist trying to earn your place at some University... but otherwise totally unusable in production environments.