FruitNinja: 3D Object Interior Texture Generation with Gaussian Splatting

Abstract

In the real world, objects reveal internal textures when sliced or cut, yet this behavior is not well-studied in 3D generation tasks today. For example, slicing a virtual 3D watermelon should reveal flesh and seeds. Given that no available dataset captures an object's full internal structure and collecting data from all slices is impractical, generative methods become the obvious approach. However, current 3D generation and inpainting methods often focus on visible appearance and overlook internal textures. To bridge this gap, we introduce FruitNinja, the first method to generate internal textures for 3D objects undergoing geometric and topological changes. Our approach produces objects via 3D Gaussian Splatting (3DGS) with both surface and interior textures synthesized, enabling real-time slicing and rendering without additional optimization. FruitNinja leverages a pre-trained diffusion model to progressively inpaint cross-sectional views and applies voxel-grid-based smoothing to achieve cohesive textures throughout the object. Our OpaqueAtom GS strategy overcomes 3DGS limitations by employing densely distributed opaque Gaussians, avoiding biases toward larger particles that destabilize training and sharp color transitions for fine-grained textures. Experimental results show that FruitNinja substantially outperforms existing approaches, showcasing unmatched visual quality in real-time rendered internal views across arbitrary geometry manipulations.

Interior Textures

The video evaluation demonstrates the consistency of the generated internal textures as objects are sliced along a fixed direction. Two distinct viewing angles are provided to assess the quality of fine-grained texture details.

Motivation

NeRF: Only surface appearance

3DGS: No interior textures

Raw 3DGS: Artifacts when sliced

Method

FruitNinja generates realistic interior textures for 3D Gaussian Splatting (3DGS) objects, ensuring high fidelity during arbitrary cuts and deformations. The pipeline includes internal Gaussian filling, user-defined cross-section inpainting with a diffusion model, progressive refinement, and Opaque Atomic GS for stability.

1. Internal Gaussians Initialization

We discretize the object volume into a 3D grid and detect empty regions using an opacity field. Rays along principal axes identify low-opacity voxels, which are filled with small Gaussian primitives, laying the foundation for interior texture synthesis.

2. Conditioned Cross-section Inpainting

For each user-defined cutting plane, we render masked cross-sectional views and use a depth-conditioned diffusion model, guided by cut-specific text prompts, to inpaint realistic interior slices. These reference views anchor the internal texture learning.

3. Progressive Texture Refinement

To ensure consistency across slices, we iteratively re-optimize the reference cross-sections with SDS, jointly update 3DGS parameters, and apply voxel-grid smoothing. This harmonizes conflicting features and fills unsupervised regions.

4. Opaque Atomic 3DGS

We employ an Opaque Atomic GS configuration: particles are clipped to a small scale and assigned high opacity. This prevents unrealistic blending, stabilizes training, and preserves sharp material boundaries under slicing.

Together, these steps enable FruitNinja to synthesize robust, high-quality internal textures that remain visually coherent for real-time slicing and arbitrary geometry edits.

BibTex

@article{wu2024fruitninja3d,
  title={FruitNinja: 3D Object Interior Texture Generation with Gaussian Splatting}, 
  author={Fangyu Wu and Yuhao Chen},
  journal={arXiv preprint arXiv:2411.12089},
  year={2024}
}