search

wenqsun / DimensionX

DimensionX: Create Any 3D and 4D Scenes from a Single Image with Controllable Video Diffusion
Apache License 2.0
810 stars 48 forks source link

how to distinguish whether the generated video is 2D or 4D？ #16

Open Fancy93 opened 3 days ago

Fancy93 commented 3 days ago

Very interesting work. I have a question: how to distinguish whether the generated video is 2D or 4D in the paper?

wenqsun commented 2 days ago

Thanks for your interest! In our paper, Fig. 4 shows the spatial- and temporal-variant videos, Fig. 5 shows the 3dgs renderings of spatial-variant videos, and Fig. 6 presents the novel-view renderings of 4D scenes.

Is it clear now? Welcome to ask us if you have any questions.

Fancy93 commented 2 days ago

Thanks for your kind reply, I have another question, is the input video 2D or 4D? 1731659551392

wenqsun commented 2 days ago

Hi, the input view video means the 4D scene rendering video from the input view.

I should use clearer expression to emphasize the results in the paper. Thanks for your question.

wenqsun commented 2 days ago

I want to talk more about the input view video in Fig. 6. For the first row, we use the real-world video as the input view (front view) video and generate the 4D scene with our controllable video generation, which is used to render the novel view videos below.

Fancy93 commented 2 days ago

Oh, I see. You want to generate a scene with three perspectives, not a video, right?