BlockFusion: Expandable 3D Scene Generation using Latent Tri-plane Extrapolation

TL;DR

BlockFusion presents a scalable framework for unbounded 3D scene generation by learning diffusion in a latent tri-plane space. It converts training blocks to raw tri-planes, compresses them with a VAE, and performs diffusion over latent tri-planes, enabling high-quality, diverse shapes. The method supports scene expansion via latent-tri-plane extrapolation conditioned by overlapping regions and 2D layout maps, followed by non-rigid surface refinement to reduce seams. Experimental results across indoor and outdoor scenarios show superior geometric fidelity, layout-controlled generation, and the ability to construct large, unbounded scenes. This approach offers a practical, high-fidelity path toward open-world 3D content generation with potential applications in games and virtual production.

Abstract

We present BlockFusion, a diffusion-based model that generates 3D scenes as unit blocks and seamlessly incorporates new blocks to extend the scene. BlockFusion is trained using datasets of 3D blocks that are randomly cropped from complete 3D scene meshes. Through per-block fitting, all training blocks are converted into the hybrid neural fields: with a tri-plane containing the geometry features, followed by a Multi-layer Perceptron (MLP) for decoding the signed distance values. A variational auto-encoder is employed to compress the tri-planes into the latent tri-plane space, on which the denoising diffusion process is performed. Diffusion applied to the latent representations allows for high-quality and diverse 3D scene generation. To expand a scene during generation, one needs only to append empty blocks to overlap with the current scene and extrapolate existing latent tri-planes to populate new blocks. The extrapolation is done by conditioning the generation process with the feature samples from the overlapping tri-planes during the denoising iterations. Latent tri-plane extrapolation produces semantically and geometrically meaningful transitions that harmoniously blend with the existing scene. A 2D layout conditioning mechanism is used to control the placement and arrangement of scene elements. Experimental results indicate that BlockFusion is capable of generating diverse, geometrically consistent and unbounded large 3D scenes with unprecedented high-quality shapes in both indoor and outdoor scenarios.

Figures (16)

• Figure 1: BlockFusion training pipeline. The training contains three steps: First, 1) the training 3D blocks are converted to raw tri-planes via per-block shape fitting, c.f. Sec. \ref{['sec:trisdf']}. Then, 2) an auto-encoder compresses the raw tri-planes into a more compact latent tri-plane space, c.f. Sec. \ref{['sec:vae']}. Lastly, 3) DDPM is trained to approximate the distributions of latent tri-planes, and during this process, layout control can also be integrated, c.f. Sec. \ref{['sec:diffusion']}.
• Figure 2: Examples of randomly cropped 3D blocks.
• Figure 3: Qualitative unconditioned block generation results. NFD nfd is also based on tri-plane diffusion. They utilize occupancy value to represent shapes, whereas ours employ SDF. All three methods are trained on room blocks.
• Figure 4: 3D aware denoising U-Net. The latent tri-plane is unfolded into three independent planes to run down-sampling convolutions. After the down-sampling layers, the three feature maps are flattened into 1D tokens and concatenated together to forward through a sequence of self-attention vaswani2017attention_is_all_need and residual block by $K=6$ times. Finally, the 1D array is reshaped into planes for up-sampling convolution and reassembled into the tri-plane structure.
• Figure 5: Latent triplane extrapolation. Given the known block $P$ and the unknown block $Q$ , the goal is to extrapolate the known latent tri-plane $z^P$ to obtain the unknown tri-plane $z^Q$ (top row). This tri-plane extrapolation is factored into the extrapolation of three 2D planes separately (bottom row).