Jul 2026· Italian National Conference on Sensors· Vol 26, pp. 4612· 0 citations· 29 references
Medicine
TL;DR
SPTNet (SuperPoint Tracking Network), an efficient multi-task neural network that tightly couples feature detection, description, and dense optical flow prediction within a unified architecture, is proposed, demonstrating a highly compact footprint advantageous for prospective embedded robotic deployment.
Abstract
Robust image-to-image correspondence is a fundamental challenge for camera-based Visual Simultaneous Localization and Mapping (SLAM). Conventional approaches primarily rely on isolated local feature matching or optical flow prediction, which often suffer from limited robustness under large parallax, high computational overhead, and cumulative drift errors during long-sequence tracking. To address these limitations, we propose SPTNet (SuperPoint Tracking Network), an efficient multi-task neural network that tightly couples feature detection, description, and dense optical flow prediction within a unified architecture. The fundamental innovation of SPTNet is a Hybrid Tracking Module (HTM) governed by a novel Predictor-Corrector mechanism. Specifically, the dense optical flow field acts as a temporal prior to constrain the descriptor matching search space, while the descriptors act as a correction signal, eliminating flow-induced drift at each frame through spatially constrained Sinkhorn optimization. This synergy enables efficient feature reuse via a shared backbone, minimizing redundant computation. Comprehensive experiments on indoor and outdoor datasets demonstrate that SPTNet attains a false matching rate as low as 1.8% at a 5-pixel threshold on HPatches, substantially reduces cumulative drift on long-sequence SLAM benchmarks, and maintains a high execution speed of 35 FPS on standard GPUs, demonstrating a highly compact footprint advantageous for prospective embedded robotic deployment.
Motion-only multiobject tracking (MOT) suffers from ID switches in uniform-appearance and deformation-heavy scenes. In these settings, appearance cues become less reliable, so stable identities depend mainly on motion information. Existing methods often process all bounding-box variables together, which can weaken cues needed for prediction and matching. We address this problem by treating each trajectory as a heterogeneous multivariate time series (MTS) and redesigning the motion-only pipeline for embedding, encoding, and matching. HSMTrack separates box-state variables before modeling their temporal and cross-variable relationships, then uses deformation-aware matching for identity association. The method requires no appearance branch and can serve as a post-detection motion module in vision-sensor tracking pipelines. Its SSM-based encoder has linear complexity with respect to trajectory length, reducing modeling cost compared with attention-based alternatives. HSMTrack achieves 59.6 IDF1 and 42.9 AssA on DanceTrack, and 77.9 IDF1 and 67.2 AssA on SportsMOT. Under a unified end-to-end protocol, it reaches 34.1 frames/s on RTX 4090 and 10.5 frames/s on Jetson Orin NX.
Tianjing Cheng, Qingyuan Yu, Bo Jiang et al.· IEEE Sensors Journal· 0 citations
This work introduces a system that combines the strong sequential constraints of SLAM with the flexibility and global optimization of offline SfM, enabling the metric reconstruction of arbitrary, long, uncalibrated videos.
Zador Pataki, Paul-Edouard Sarlin, Marc Pollefeys· 0 citations
Monocular RGB dense SLAM remains challenging because depth and global metric scale are not directly observable from a single camera. Existing systems often combine optical-flow and monocular-geometry priors predicted by independently trained networks, which can provide inconsistent constraints to bundle adjustment (BA). Our quantitative prior-consistency analysis indicates that this disagreement is an important contributor to unstable local optimization and reconstruction error rather than the sole cause of drift. We propose KP-SLAM, which predicts dense optical flow and paired pointmap priors from a shared representation and incorporates them into the same BA backend. We further introduce a Depth-Scale-Pose-to-Pointmap (DSPP) objective that relates optimized inverse depth, edge-wise relative scale, and camera pose to paired pointmap constraints. Experiments on ScanNet, TUM-RGBD, KITTI, Tanks-and-Temples, and dynamic sequences show improved tracking, depth, and rendering metrics over the compared RGB-only baselines under the reported settings. The results support the usefulness of synchronized priors while also revealing remaining limitations in highly dynamic, weakly textured, and large-scale scenes.
Song Gao, Xinyu Huang, Zheng Huang et al.· Symmetry· 0 citations
Matchmoving is the bedrock of visual effects, yet it remains a fragile bottleneck when footage contains heavy motion blur, low texture, or dynamic occlusion. While physical on-set camera tracking (e.g., encoded cranes) exists, it is often impractical for handheld interior shots and prone to mechanical slippage, leaving post-production software to solve the gap. This talk presents a LiDAR-Constrained Deep Visual Odometry system, a robust tracking architecture designed to solve these "impossible" shots by fusing pre-existing LiDAR geometry with modern Deep Learning. Unlike traditional commercial solvers that hunt for sparse, high-contrast corners, our approach uses Deep Optical Flow (RAFT) to track the entire dense image context, locking the camera directly to the set’s 3D mesh. We introduce a "Leapfrogging" architecture that automatically detects and corrects temporal drift by re-anchoring to the geometry from trusted keyframes. By prioritizing geometric truth over feature quantity, this standalone Python tool reduces days of manual hand-tracking and rotoscoping to minutes of automated computation, achieving high median precision on sequences where standard algorithms fail entirely.
Srdjan Milosevic· Proceedings of the Special I...· 0 citations
Neural Radiance Fields (NeRF)-based SLAM has demonstrated impressive results in small-scale scene reconstruction, yet scaling these methods to extensive, complex environments remains challenging due to catastrophic forgetting and accumulated trajectory drift. This paper presents a robust, large-scale neural SLAM system featuring a multi-submap architecture and a dual-tier loop closure mechanism. Specifically, we propose a progressive mapping strategy that dynamically allocates neural submaps to maintain high-fidelity representations without memory explosion. For robust pose estimation, an optical-flow-based tracking module is integrated to handle aggressive motions. To address global consistency, we introduce a local-to-global loop closure framework leveraging the foundation model for high-performance global descriptor extraction, significantly enhancing relocalization accuracy under varying viewpoints. Furthermore, an inter-submap online distillation algorithm is designed during back-end optimization to enforce geometric and appearance consistency across overlapping submap boundaries. To validate the system, we developed a customized handheld mechatronic platform and conducted extensive evaluations on both public benchmarks and our large-scale indoor-outdoor datasets. Experimental results, including direct deployment on an onboard computing unit, demonstrate that our approach outperforms state-of-the-art neural SLAM methods in reconstruction quality and localization robustness, providing a scalable solution for real-world robotic perception and digital twinning. We will release the code publicly on \href{https://github.com/dtc111111/MSN-SLAM}{https://github.com/dtc111111/MSN-SLAM} .
Tianchen Deng, Chongdi Wang, Nailin Wang et al.· 0 citations
Many feature-based visual-inertial odometry (VIO) systems rely on sparse feature tracking, whose accuracy and robustness directly affect state estimation. Classical KLT trackers rely primarily on local image patches and can become unreliable under rapid motion or in low-texture environments. We propose KLTNet, a lightweight learning-based, plug-and-play sparse feature tracker designed to replace classical KLT trackers in KLT-based VIO front ends. KLTNet follows a coarse-to-fine, dense-to-sparse architecture that combines low-resolution dense optical flow for robust global motion initialization with triplet-patch refinement for accurate and temporally consistent tracking. A fixed reference patch provides a stable anchor throughout each feature track and helps reduce accumulated tracking drift. In addition, KLTNet predicts anisotropic confidence weights supervised through differentiable multi-view triangulation, which can be used as observation weights in compatible VIO estimators. Experiments with VINS-Mono and OpenVINS on public benchmarks and a self-collected low-texture dataset demonstrate improved tracking and odometry accuracy over classical KLT, while maintaining real-time performance on an embedded platform.