Fluid-SDF: Ultra-Lightweight and Editable Implicit Shape Representation via Differentiable Primitives
Quick Answer
Fluid-SDF introduces a differentiable CSG framework that models complex shapes with under 100 parameters, outperforming traditional INRs in noise resistance and enabling direct user editing without retraining.
Quick Take
This makes it ideal for mobile AI and resource-constrained environments.
Key Points
- Fluid-SDF uses under 100 parameters for complex shape modeling.
- Outperforms standard neural baselines in intersection-over-union (mIoU).
- Resists high-frequency dataset noise better than capacity-matched networks.
- Enables zero-shot user editing of shape features without retraining.
- Optimized for mobile AI and augmented reality applications.
DeepSignal Analysis
What happened
Fluid-SDF is a new framework for modeling shapes using fewer than 100 parameters, which improves upon traditional implicit neural representations (INRs) by enhancing noise resistance and allowing for direct user editing. This framework is particularly beneficial for applications in mobile AI and environments with limited resources.
Key evidence
- Fluid-SDF uses a differentiable Constructive Solid Geometry (CSG) framework that models shapes with under 100 parameters, which is a significant reduction compared to traditional methods.
- The framework demonstrates improved resistance to high-frequency dataset noise, where conventional neural networks tend to overfit, making it more reliable for practical applications.
- Fluid-SDF allows for zero-shot user editing of shape features without the need for retraining, which is a notable advantage over standard INRs.
Why it matters
The introduction of Fluid-SDF addresses critical limitations of existing shape modeling techniques, particularly in terms of editability and noise resistance. By enabling efficient shape representation with fewer parameters, it opens up new possibilities for deployment in mobile and resource-constrained environments. This could lead to advancements in augmented reality and other applications where computational resources are limited.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Implicit Neural Representations (INRs) have become the standard for continuous 2D shape modeling, but they suffer from black-box uneditability, vulnerability to noise, and high parameter counts that severely hinder deployment on edge devices. We introduce Fluid-SDF, a highly compressed, differentiable Constructive Solid Geometry (CSG) framework that models shapes using explicit geometric primitives blended via a smooth minimum function. By replacing traditional multi-layer perceptrons (MLPs) with a parameterized primitive engine, Fluid-SDF reconstructs complex, non-convex topologies using strictly under 100 parameters, achieving comparable or superior intersection-over-union (mIoU) to standard neural baselines. Furthermore, we demonstrate that Fluid-SDF acts as a powerful geometric prior, inherently resisting high-frequency dataset noise where capacity-matched neural networks catastrophically overfit. Finally, unlike standard INRs, Fluid-SDF's explicit parameter space allows for direct, zero-shot user editing of local and global shape features without retraining. By bypassing expensive on-device gradient updates entirely, Fluid-SDF is uniquely suited for mobile AI, augmented reality, and resource-constrained embedded environments
| Comments: | 6 pages, 5 figures |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2607.18646 [cs.CV] |
| (or arXiv:2607.18646v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.18646 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Pradyumna Sripada [view email]
[v1]
Tue, 21 Jul 2026 02:35:20 UTC (367 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.