OREO is a framework for improving the visual fidelity of 3D generation. It first reinforces rendered views using image editing, then uses a latent contrastive objective to distill the gap between the edited and original views back into the training generator. This on-the-fly optimization loop aims to align the generated 3D content with higher-quality 2D editing results without requiring additional training data.
Dynamic optimization loop aligns 3D generators with 2D diffusion priors to improve realism
OREO uses on-the-fly rendered-view editing and reinforced learning to refine 3D assets from pre-trained models without static datasets.
Chinese Tech
The Hong Kong Polytechnic University · Tencent ARC Lab
Research Digest··2 min read
In one line
OREO enhances 3D generation fidelity by editing rendered views and distilling the gap back into the generator via latent contrastive learning.
Before you read this
Three ideas this paper builds ondiffusion modelA type of generative model that learns to reverse a gradual noising process to create data.197 papers →flow matchingA generative training method that learns an ordinary differential equation mapping noise to data by matching velocity fields along probability paths.173 papers →contrastive alignmentA method that learns embeddings by minimizing distance between positive pairs and maximizing distance between negative pairs.48 papers →
Read it as
The authors propose OREO, a framework that enhances visual fidelity of 3D generators by creating a dynamic optimization loop.
Ma et al. introduce OREO, a fidelity alignment framework for 3D generation.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.