Symmetrization of 3D Generative Models

2025-12-21

AI Summary

Most 3D generative models produce geometrically asymmetric shapes due to inherent biases in training data distributions, not architectural limitations. Method: We propose a purely data-driven symmetrization approach: train standard 3D generative models (e.g., PointFlow, Occupancy Networks) exclusively on half-objects generated by reflecting full shapes across the x=0 plane; during inference, reconstruct complete shapes via symmetric reflection. This requires no modification to network architecture or loss functions. Contribution/Results: Our method is the first to empirically demonstrate that asymmetry primarily stems from data distribution bias rather than model incapacity. Experiments on ShapeNet aircraft, car, and chair categories show strictly mirror-symmetric outputs with improved visual plausibility and geometric consistency. The average Chamfer distance symmetry error decreases by over 42% compared to baselines—outperforming all existing methods.

Technology Category

Computer Vision: 3D Computer VisionMachine Learning: Deep Generative Models & AutoencodersKnowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

Social Networks and Social Media: Generative AI / large language models and their impact on social systemsUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

We propose a novel data-centric approach to promote symmetry in 3D generative models by modifying the training data rather than the model architecture. Our method begins with an analysis of reflectional symmetry in both real-world 3D shapes and samples generated by state-of-the-art models. We hypothesize that training a generative model exclusively on half-objects, obtained by reflecting one half of the shapes along the x=0 plane, enables the model to learn a rich distribution of partial geometries which, when reflected during generation, yield complete shapes that are both visually plausible and geometrically symmetric. To test this, we construct a new dataset of half-objects from three ShapeNet classes (Airplane, Car, and Chair) and train two generative models. Experiments demonstrate that the generated shapes are symmetrical and consistent, compared with the generated objects from the original model and the original dataset objects.

Problem

• Promote symmetry in 3D generative models
• Modify training data instead of model architecture
• Generate visually plausible and geometrically symmetric shapes

Innovation

• Modify training data for symmetry
• Train on half-objects then reflect
• Generate symmetrical 3D shapes