Robust Single-Stage Fully Sparse 3D Object Detection via Detachable Latent Diffusion

2025-08-05

AI Summary

Existing 3D detection methods based on Denoising Diffusion Probabilistic Models (DDPMs) suffer from multi-step iterative inference, low efficiency, and insufficient robustness. To address these limitations, this paper proposes RSDNet, a single-stage fully sparse detection framework. Our key contributions are: (1) a separable latent diffusion mechanism enabling one-step denoising across hierarchical sparse feature spaces; (2) a semantic-geometric joint conditional guidance scheme that mitigates the absence of center-point features in sparse representations and enhances robustness against diverse and multi-level noise; and (3) a lightweight multi-level denoising autoencoder integrated into an end-to-end reconstruction pipeline. Evaluated on mainstream benchmarks, RSDNet achieves state-of-the-art detection performance with significantly reduced inference overhead, striking an optimal balance between high efficiency and strong robustness.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Denoising Diffusion Probabilistic Models (DDPMs) have shown success in robust 3D object detection tasks. Existing methods often rely on the score matching from 3D boxes or pre-trained diffusion priors. However, they typically require multi-step iterations in inference, which limits efficiency. To address this, we propose a extbf{R}obust single-stage fully extbf{S}parse 3D object extbf{D}etection extbf{Net}work with a Detachable Latent Framework (DLF) of DDPMs, named RSDNet. Specifically, RSDNet learns the denoising process in latent feature spaces through lightweight denoising networks like multi-level denoising autoencoders (DAEs). This enables RSDNet to effectively understand scene distributions under multi-level perturbations, achieving robust and reliable detection. Meanwhile, we reformulate the noising and denoising mechanisms of DDPMs, enabling DLF to construct multi-type and multi-level noise samples and targets, enhancing RSDNet robustness to multiple perturbations. Furthermore, a semantic-geometric conditional guidance is introduced to perceive the object boundaries and shapes, alleviating the center feature missing problem in sparse representations, enabling RSDNet to perform in a fully sparse detection pipeline. Moreover, the detachable denoising network design of DLF enables RSDNet to perform single-step detection in inference, further enhancing detection efficiency. Extensive experiments on public benchmarks show that RSDNet can outperform existing methods, achieving state-of-the-art detection.

Problem

• Improves efficiency in 3D object detection with single-step inference
• Enhances robustness to multi-level perturbations via denoising networks
• Addresses sparse representation issues with semantic-geometric guidance

Innovation

• Detachable Latent Framework for efficient denoising
• Semantic-geometric guidance for object perception
• Single-step detection for enhanced efficiency