SeMo3D unifies 3D generation editing and completion
Original title: SeMo3D: Unifying 3D Generation, Editing, and Completion via Selective Modeling
SeMo3D uses selective modeling to unify 3D tasks with region-selective denoising and Omni3DiT architecture.
SeMo3D addresses task isolation in 3D generation editing and completion by selective modeling that keeps unchanged content intact by construction. Region-selective denoising randomly selects regions of varying sizes on standard 3D assets and applies noise and loss only to them, enabling training without paired editing or completion data. Omni3DiT is a multi-condition 3D diffusion transformer that distinguishes regions to create from those to retain and fuses image text and point guidance for fine-grained control. Experiments across three benchmarks show SeMo3D surpasses state-of-the-art methods on all tasks and serves as a unified modeling tool for agents.
- Unifies generation editing and completion via selective modeling
- Region-selective denoising for training
- Omni3DiT architecture with multi-condition fusion
- Strong consistency and accuracy in editing and completion
- Serves as tool for agents through natural-language interaction
From the original article
1 min read at huggingface.co
3D generation, editing, and completion are typically treated as regenerating the complete 3D asset under task-specific constraints, even when only a local region needs modification. This leads to task isolation, inconsistency in content that should remain unchanged, and dependence on task-specific paired data.
By Xiaokun Sun, Jian Yang, Zhenyu Zhang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.