Albedo map is necessary？

[Reuben-Sun]

I find the inference code need both obj model, albedo map and text prompt. But I don't understand why this task should input a good albedo map.
Is there any way to generate a UV base texture without using the albedo texture, relying solely on the model.obj and text prompt? Or I have a front view render image and obj model, can I generate the texture?

[XinYu-Andy]

Our method relies on having a posed image to guide texture generation. In other words, we need a known pose when synthesizing the texture. For the evaluation purpose, we use an existing texture map to render a view, which in turn provides accurate pose information; we are not estimating the pose from an arbitrary image. While generating texture directly from unposed images might be more practical in real-world scenarios, that goes beyond the scope of what our current study aims to address.