TIMI: Training-Free Image-to-3D Multi-Instance Generation with Spatial Fidelity

2026-03-01

AI Summary

Existing image-to-3D multi-instance generation methods struggle to preserve spatial fidelity and often rely on fine-tuning, resulting in high training costs and limited performance. This work proposes a novel framework that achieves high-quality multi-instance 3D generation without any additional training, leveraging pre-trained image-to-3D (I23D) models for the first time. By introducing Instance-aware Separation Guidance (ISG) and Spatially Stable Geometric Update (SGU) mechanisms, the method effectively disentangles instance-level ambiguities while preserving the relative spatial relationships among instances. The approach outperforms existing methods in both global layout coherence and local geometric detail, while also enabling faster inference. Consequently, it significantly enhances spatial fidelity and computational efficiency in multi-instance 3D generation.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: 3D Computer Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Precise spatial fidelity in Image-to-3D multi-instance generation is critical for downstream real-world applications. Recent work attempts to address this by fine-tuning pre-trained Image-to-3D (I23D) models on multi-instance datasets, which incurs substantial training overhead and struggles to guarantee spatial fidelity. In fact, we observe that pre-trained I23D models already possess meaningful spatial priors, which remain underutilized as evidenced by instance entanglement issues. Motivated by this, we propose TIMI, a novel Training-free framework for Image-to-3D Multi-Instance generation that achieves high spatial fidelity. Specifically, we first introduce an Instance-aware Separation Guidance (ISG) module, which facilitates instance disentanglement during the early denoising stage. Next, to stabilize the guidance introduced by ISG, we devise a Spatial-stabilized Geometry-adaptive Update (SGU) module that promotes the preservation of the geometric characteristics of instances while maintaining their relative relationships. Extensive experiments demonstrate that our method yields better performance in terms of both global layout and distinct local instances compared to existing multi-instance methods, without requiring additional training and with faster inference speed.

Problem

• Image-to-3D
• multi-instance generation
• spatial fidelity
• training-free
• instance disentanglement

Innovation

• Training-free
• Image-to-3D
• Multi-instance generation
• Spatial fidelity
• Instance disentanglement