2026· ITM Web of Conferences· Vol 89, pp. 01003· 0 citations
TL;DR
A depth-guided lightweight multi-task 3D scene understanding method based on a lightweight U-Net architecture, uses RGB and depth four-channel dual- modal input, and constructs a network structure with a shared encoder and independent decoders, allowing a single inference to simultaneously accomplish the three core tasks of semantic segmentation, depth completion, and obstacle detection.
Abstract
To address the core issues faced by edge-side 3D scene perception of mobile robots, such as the difficulty of balancing accuracy and real-time performance, insufficient multi-task fusion, and deviations in depth geometric consistency, this paper proposes a depth-guided lightweight multi-task 3D scene understanding method. This method is based on a lightweight U-Net architecture, uses RGB and depth four-channel dual- modal input, and constructs a network structure with a shared encoder and independent decoders, allowing a single inference to simultaneously accomplish the three core tasks of semantic segmentation, depth completion, and obstacle detection. It also Introduce depth consistency error (DCE) to construct a weighted joint loss function, strengthening scene geometry structure learning; through a dual lightweight strategy of structured pruning and INT8 quantization, precisely adapt to NVIDIA Jetson Xavier NX edge hardware. Experimental results on the NYU Depth V2 and KITTI datasets show that this method achieves a semantic segmentation mIoU ≥ 60%, depth completion RMSE ≤ 1.0m, obstacle detection F1-Score ≥ 85%, and edge- end inference speed ≥ 15 FPS, effectively balancing perception accuracy and real-time performance, providing a highly practical lightweight solution for autonomous perception in mobile robots.
A 3D Local- Global Linear Attention Mechanism (LG-LAM) is devised that efficiently captures long-range contextual information with linear complexity, enabling a comprehensive understanding of the 3D scene without heavy computational burdens.
Jie Li, Jiaheng Xu, Laiyan Ding et al.· International Conference on...· 0 citations
In recent years, Transformer-based 3D vision foundation models have demonstrated strong generalization in multiview geometry and scene reconstruction. However, their scalability to long-sequence, urban-scale RGB streams remains limited due to the quadratic complexity of attention mechanisms, redundant frame processing, and high GPU memory pressure caused by dense spatial tokens. Although VGGT-Long partially alleviates these issues through chunked inference and loop-closure optimization, its geometric reasoning pipeline still involves substantial redundant computation, making it difficult to balance efficiency and accuracy in long-sequence scenarios. In this paper, we revisit the computational bottlenecks in long-sequence geometric inference, with a particular focus on spatial token redundancy and cross-frame global attention. We propose π³-LEGS, an efficient geometric inference system designed for scalable long-sequence 3D reconstruction. π³-LEGS integrates three key components: (1) a π³-based permutation-equivariant inference module to enhance unordered multi-view feature aggregation; (2) a geometry-aware keyframe selection mechanism that dynamically filters low-contribution frames to reduce redundant computation; and (3) a training-free block-sparse attention strategy that adaptively generates sparse attention masks based on pooled Query–Key similarity, significantly reducing global attention overhead. Extensive experiments on the KITTI Odometry dataset demonstrate that π³-LEGS achieves an average Absolute Trajectory Error (ATE) of 26.51, improving upon VGGT-Long by 6.4%, while reducing end-to-end inference time by 15.9%. Moreover, the proposed system maintains stable performance on thousand-frame sequences without runtime failures, highlighting its effectiveness in achieving a favorable accuracy efficiency trade-off for large-scale 3D reconstruction.
Faline Fu, Xiaoli Cao, Can Tang et al.· International Conference on...· 0 citations
Self-supervised monocular depth estimation has emerged as an appealing solution to design lightweight and effective models for deployment on computationally constrained devices due to its reduced reliance on expensive depth sensors. By eliminating the need for ground-truth annotations and leveraging the simplicity of monocular camera setups, this approach facilitates cost-effective data collection and broad applicability across fields such as computer vision and robotics. A critical challenge is achieving resource-efficient neural networks without compromising the overall performance. State-of-the-art models generally adopt depth-wise convolutions and attention mechanisms; however, these functions often incur high energy costs and face compatibility issues in embedded environments. To address this, we propose XiDepth, a lightweight architecture based on the XiNet operator block, designed to enhance feature extraction while maintaining low computational complexity and energy demand. On the KITTI dataset, XiDepth achieves state-of-the-art performance with only 0.8M parameters. Tests on a Raspberry Pi 4 further confirm its suitability for real-world embedded applications, reducing FLOPs by 40% and energy consumption by 35% compared to leading methods.
Elena Izzo, Riccardo Toniolo, Lamberto Ballan· 1 citation
Semantically-guided progressive network (SGP-Net) is proposed, a semantically guided progressive refinement framework for MDE based on multi-task learning that improves key relative-error and accuracy metrics over the DCDepth baseline and remains competitive with recent methods.
Henan Hu, Xu Cheng, Rong-Hua Li et al.· Measurement science and tech...· 0 citations
Camera-based 3D semantic scene completion (SSC) provides comprehensive scene understanding for autonomous driving and robotics. However, existing methods often treat stereo depth estimates as deterministic geometric constraints, causing depth uncertainty and local correspondence errors to propagate directly into voxel representations. To address this issue, we propose RayLift, a framework that uses stereo geometry as a metric reference while incorporating complementary ray evidence to recover reliable 3D structures adaptively. RayLift first employs a Complementary Context Encoder that extracts geometry-aware priors from a frozen 3D vision foundation model, thereby enriching the scene context. It then introduces a Depth Ray Evidence Lifter module that jointly models geometric dissimilarity, depth confidence, and spatial uncertainty to adaptively sample and weight candidate surface locations along each camera ray. Finally, a Semantic-Aware Voxel Integrator injects the resulting ray evidence into voxel features by explicitly modeling their spatial support. Extensive experiments on SemanticKITTI and SSCBench-KITTI-360 demonstrate that RayLift achieves competitive performance and consistently outperforms existing methods.
Meng Wang, Hongxia Yu, Wenzhe He et al.· 0 citations
Monocular depth estimation is an important dense prediction task for autonomous driving, robotic perception, and unmanned aerial vehicle navigation. Although recent deep networks have achieved impressive accuracy, many of them depend on large backbones and expensive context modeling modules, making deployment on resource-constrained platforms difficult. This paper proposes a lightweight supervised monocular depth estimation network that improves feature representation while maintaining a compact model size. The proposed framework adopts an efficient encoder and applies Dilated-Strip Enhancement blocks to multi-scale encoder features. By combining dilated depthwise convolution with horizontal and vertical strip convolutions, the proposed block enlarges the receptive field and captures directional spatial structures efficiently. A multi-scale fusion decoder further integrates enhanced features from different resolutions, and auxiliary supervision is used to guide intermediate depth representations. Experiments on NYU Depth V2 and KITTI show that the proposed model achieves competitive accuracy with only 4.2M parameters. In particular, the proposed method achieves strong performance on the KITTI benchmark, suggesting that the dilated-strip design is effective for outdoor scenes with elongated and directional structures.
Dun-Yu Hu, Han-Xiang Zhang, J. Gu et al.· 2026 IEEE Canadian Atlantic...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
