StripDet: Strip Attention-Based Lightweight 3D Object Detection from Point Cloud

2025-09-07

AI Summary

To address computational and memory bottlenecks in deploying point-cloud 3D object detection models on edge devices, this paper proposes StripDet, a lightweight real-time detection framework. Methodologically, StripDet introduces three key innovations: (1) the Strip Attention Block, which decomposes 2D convolutions into asymmetric stripe convolutions to linearly reduce computational complexity; (2) a hardware-friendly multi-scale backbone network integrating stripe convolutions, depthwise separable convolutions, and lightweight attention mechanisms; and (3) an end-to-end multi-scale feature fusion strategy to enhance long-range spatial modeling. Evaluated on the KITTI dataset, StripDet achieves 79.97% mAP with only 0.65M parameters—seven times fewer than PointPillars—while outperforming existing lightweight and knowledge distillation approaches. The framework demonstrates superior efficiency–accuracy trade-offs for resource-constrained edge deployment.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Learning & Optimization for CV
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

The deployment of high-accuracy 3D object detection models from point cloud remains a significant challenge due to their substantial computational and memory requirements. To address this, we introduce StripDet, a novel lightweight framework designed for on-device efficiency. First, we propose the novel Strip Attention Block (SAB), a highly efficient module designed to capture long-range spatial dependencies. By decomposing standard 2D convolutions into asymmetric strip convolutions, SAB efficiently extracts directional features while reducing computational complexity from quadratic to linear. Second, we design a hardware-friendly hierarchical backbone that integrates SAB with depthwise separable convolutions and a simple multiscale fusion strategy, achieving end-to-end efficiency. Extensive experiments on the KITTI dataset validate StripDet's superiority. With only 0.65M parameters, our model achieves a 79.97% mAP for car detection, surpassing the baseline PointPillars with a 7x parameter reduction. Furthermore, StripDet outperforms recent lightweight and knowledge distillation-based methods, achieving a superior accuracy-efficiency trade-off while establishing itself as a practical solution for real-world 3D detection on edge devices.

Problem

• Lightweight 3D object detection from point clouds
• Reducing computational and memory requirements
• Efficient on-device deployment for edge devices

Innovation

• Strip Attention Block captures long-range dependencies
• Asymmetric strip convolutions reduce computational complexity
• Hardware-friendly backbone integrates depthwise separable convolutions