We introduce iShapEditing, a novel framework for 3D shape editing which is applicable to both generated and real shapes. Users manipulate shapes by dragging handle points to corresponding targets, offering an intuitive and intelligent editing interface.
Leveraging the Triplane Diffusion model and robust intermediate feature correspondence, our framework utilizes classifier guidance to adjust noise representations during sampling process, ensuring alignment with user expectations while preserving plausibility. For real shapes, we employ shape predictions at each time step alongside a DDPM-based inversion algorithm to derive their latent codes, facilitating seamless editing.
iShapEditing provides effective and intelligent control over shapes without the need for additional model training or fine-tuning. Experimental examples demonstrate the effectiveness and superiority of our method in terms of editing accuracy and plausibility.
Users simply need to select certain handle points(marked in red) and their target points (marked in blue). Our method accurately moves these handle points to reach their corresponding targets while maintaining shape plausibility.
For a user-supplied watertight shape, we first guide the generation of an ``in-distribution'' triplane representation and employ a DDPM-based inversion algorithm to derive its latent code. Then we perform the dragging process.
For distant target points, we guide handle points towards their targets step-by-step.
Our approach can be extended to a wider range of shape editing operations, such as copy and delete.
@inproceedings{li2024ishapediting,
title={iShapEditing: Intelligent Shape Editing with Diffusion Models},
author={Li, Jing and Zhang, Juyong and Chen, Falai},
booktitle={Computer Graphics Forum},
pages={e15253},
year={2024},
organization={Wiley Online Library}
}