Danping Zou

Xinjiang Wang, Zeyu Liu, Yu Hu et al.

We introduce a lightweight network to improve descriptors of keypoints within the same image. The network takes the original descriptors and the geometric properties of keypoints as the input, and uses an MLP-based self-boosting stage and a Transformer-based cross-boosting stage to enhance the descriptors. The boosted descriptors can be either real-valued or binary ones. We use the proposed network to boost both hand-crafted (ORB, SIFT) and the state-of-the-art learning-based descriptors (SuperPoint, ALIKE) and evaluate them on image matching, visual localization, and structure-from-motion tasks. The results show that our method significantly improves the performance of each task, particularly in challenging cases such as large illumination changes or repetitive patterns. Our method requires only 3.2ms on desktop GPU and 27ms on embedded GPU to process 2000 features, which is fast enough to be applied to a practical system. The code and trained weights are publicly available at github.com/SJTU-ViSYS/FeatureBooster.

Fanxing Li, Shengyang Wang, Fangyu Sun et al.

Existing FPV object tracking methods heavily rely on handcrafted modular pipelines, which incur high onboard computation and cumulative errors. While learning-based approaches have mitigated computational delays, most still generate only high-level trajectories (position and yaw). This loose coupling with a separate controller sacrifices precise attitude control; consequently, even if target is localized precisely, accurate target estimation does not ensure that the body-fixed camera is consistently oriented toward the target, it still probably degrades and loses target when tracking high-maneuvering target. To address these challenges, we present StableTracker, a learning-based control policy that enables quadrotors to robustly follow a moving target from arbitrary viewpoints. The policy is trained using backpropagation-through-time via differentiable simulation, allowing the quadrotor to keep a fixed relative distance while maintaining the target at the center of the visual field in both horizontal and vertical directions, thereby functioning as an autonomous aerial camera. We compare StableTracker against state-of-the-art traditional algorithms and learning baselines. Simulation results demonstrate superior accuracy, stability, and generalization across varying safe distances, trajectories, and target velocities. Furthermore, real-world experiments on a quadrotor with an onboard computer validate the practicality of the proposed approach.

Zinan Lv, Yeqian Qian, Chen Sang et al.

UAV navigation in unstructured outdoor environments using passive monocular vision is hindered by the substantial visual domain gap between simulation and reality. While 3D Gaussian Splatting enables photorealistic scene reconstruction from real-world data, existing methods inherently couple static lighting with geometry, severely limiting policy generalization to dynamic real-world illumination. In this paper, we propose a novel end-to-end reinforcement learning framework designed for effective zero-shot transfer to unstructured outdoors. Within a high-fidelity simulation grounded in real-world data, our policy is trained to map raw monocular RGB observations directly to continuous control commands. To overcome photometric limitations, we introduce Relightable 3D Gaussian Splatting, which decomposes scene components to enable explicit, physically grounded editing of environmental lighting within the neural representation. By augmenting training with diverse synthesized lighting conditions ranging from strong directional sunlight to diffuse overcast skies, we compel the policy to learn robust, illumination-invariant visual features. Extensive real-world experiments demonstrate that a lightweight quadrotor achieves robust, collision-free navigation in complex forest environments at speeds up to 10 m/s, exhibiting significant resilience to drastic lighting variations without fine-tuning.

Ang Li, Anning Hu, Wei Xi et al.

Accurate and dense depth estimation with stereo cameras and LiDAR is an important task for automatic driving and robotic perception. While sparse hints from LiDAR points have improved cost aggregation in stereo matching, their effectiveness is limited by the low density and non-uniform distribution. To address this issue, we propose a novel stereo-LiDAR depth estimation network with Semi-Dense hint Guidance, named SDG-Depth. Our network includes a deformable propagation module for generating a semi-dense hint map and a confidence map by propagating sparse hints using a learned deformable window. These maps then guide cost aggregation in stereo matching. To reduce the triangulation error in depth recovery from disparity, especially in distant regions, we introduce a disparity-depth conversion module. Our method is both accurate and efficient. The experimental results on benchmark tests show its superior performance. Our code is available at https://github.com/SJTU-ViSYS/SDG-Depth.

Fanxing Li, Fangyu Sun, Tianbao Zhang et al.

Quadrotor control policies can be trained with high performance using the exact gradients of the rewards to directly optimize policy parameters via backpropagation-through-time (BPTT). However, designing a fully differentiable reward architecture is often challenging. Partially differentiable rewards will result in biased gradient propagation that degrades training performance. To overcome this limitation, we propose Amended Backpropagation-through-Time (ABPT), a novel approach that mitigates gradient bias while preserving the training efficiency of BPTT. ABPT combines 0-step and N-step returns, effectively reducing the bias by leveraging value gradients from the learned Q-value function. Additionally, it adopts entropy regularization and state initialization mechanisms to encourage exploration during training. We evaluate ABPT on four representative quadrotor flight tasks \li{in both real world and simulation}. Experimental results demonstrate that ABPT converges significantly faster and achieves higher ultimate rewards than existing learning algorithms, particularly in tasks involving partially differentiable rewards. The code will be released at http://github.com/Fanxing-LI/ABPT.

Ziken Huang, Xinze Niu, Bowen Chai et al.

High-speed aerial grasping presents significant challenges due to the high demands on precise, responsive flight control and coordinated gripper manipulation. In this work, we propose Swooper, a deep reinforcement learning (DRL) based approach that achieves both precise flight control and active gripper control using a single lightweight neural network policy. Training such a policy directly via DRL is nontrivial due to the complexity of coordinating flight and grasping. To address this, we adopt a two-stage learning strategy: we first pre-train a flight control policy, and then fine-tune it to acquire grasping skills. With the carefully designed reward functions and training framework, the entire training process completes in under 60 minutes on a standard desktop with an Nvidia RTX 3060 GPU. To validate the trained policy in the real world, we develop a lightweight quadrotor grasping platform equipped with a simple off-the-shelf gripper, and deploy the policy in a zero-shot manner on the onboard Raspberry Pi 4B computer, where each inference takes only about 1.0 ms. In 25 real-world trials, our policy achieves an 84% grasp success rate and grasping speeds of up to 1.5 m/s without any fine-tuning. This matches the robustness and agility of state-of-the-art classical systems with sophisticated grippers, highlighting the capability of DRL for learning a robust control policy that seamlessly integrates high-speed flight and grasping. The supplementary video is available for more results. Video: https://zikenhuang.github.io/Swooper/.

Boying Li, Danping Zou, Yuan Huang et al.

We propose a novel visual SLAM method that integrates text objects tightly by treating them as semantic features via fully exploring their geometric and semantic prior. The text object is modeled as a texture-rich planar patch whose semantic meaning is extracted and updated on the fly for better data association. With the full exploration of locally planar characteristics and semantic meaning of text objects, the SLAM system becomes more accurate and robust even under challenging conditions such as image blurring, large viewpoint changes, and significant illumination variations (day and night). We tested our method in various scenes with the ground truth data. The results show that integrating texture features leads to a more superior SLAM system that can match images across day and night. The reconstructed semantic 3D text map could be useful for navigation and scene understanding in robotic and mixed reality applications. Our project page: https://github.com/SJTU-ViSYS/TextSLAM .

Jie Yin, Ang Li, Tao Li et al.

We introduce M2DGR: a novel large-scale dataset collected by a ground robot with a full sensor-suite including six fish-eye and one sky-pointing RGB cameras, an infrared camera, an event camera, a Visual-Inertial Sensor (VI-sensor), an inertial measurement unit (IMU), a LiDAR, a consumer-grade Global Navigation Satellite System (GNSS) receiver and a GNSS-IMU navigation system with real-time kinematic (RTK) signals. All those sensors were well-calibrated and synchronized, and their data were recorded simultaneously. The ground truth trajectories were obtained by the motion capture device, a laser 3D tracker, and an RTK receiver. The dataset comprises 36 sequences (about 1TB) captured in diverse scenarios including both indoor and outdoor environments. We evaluate state-of-the-art SLAM algorithms on M2DGR. Results show that existing solutions perform poorly in some scenarios. For the benefit of the research community, we make the dataset and tools public. The webpage of our project is https://github.com/SJTU-ViSYS/M2DGR.

Boying Li, Yuan Huang, Zeyu Liu et al.

Self-supervised monocular depth estimation has achieved impressive performance on outdoor datasets. Its performance however degrades notably in indoor environments because of the lack of textures. Without rich textures, the photometric consistency is too weak to train a good depth network. Inspired by the early works on indoor modeling, we leverage the structural regularities exhibited in indoor scenes, to train a better depth network. Specifically, we adopt two extra supervisory signals for self-supervised training: 1) the Manhattan normal constraint and 2) the co-planar constraint. The Manhattan normal constraint enforces the major surfaces (the floor, ceiling, and walls) to be aligned with dominant directions. The co-planar constraint states that the 3D points be well fitted by a plane if they are located within the same planar region. To generate the supervisory signals, we adopt two components to classify the major surface normal into dominant directions and detect the planar regions on the fly during training. As the predicted depth becomes more accurate after more training epochs, the supervisory signals also improve and in turn feedback to obtain a better depth model. Through extensive experiments on indoor benchmark datasets, the results show that our network outperforms the state-of-the-art methods. The source code is available at https://github.com/SJTU-ViSYS/StructDepth .

Jinquan Li, Ling Pei, Danping Zou et al.

This paper proposes a novel simultaneous localization and mapping (SLAM) approach, namely Attention-SLAM, which simulates human navigation mode by combining a visual saliency model (SalNavNet) with traditional monocular visual SLAM. Most SLAM methods treat all the features extracted from the images as equal importance during the optimization process. However, the salient feature points in scenes have more significant influence during the human navigation process. Therefore, we first propose a visual saliency model called SalVavNet in which we introduce a correlation module and propose an adaptive Exponential Moving Average (EMA) module. These modules mitigate the center bias to enable the saliency maps generated by SalNavNet to pay more attention to the same salient object. Moreover, the saliency maps simulate the human behavior for the refinement of SLAM results. The feature points extracted from the salient regions have greater importance in optimization process. We add semantic saliency information to the Euroc dataset to generate an open-source saliency SLAM dataset. Comprehensive test results prove that Attention-SLAM outperforms benchmarks such as Direct Sparse Odometry (DSO), ORB-SLAM, and Salient DSO in terms of efficiency, accuracy, and robustness in most test cases.

Fei Wen, Danping Zou, Rendong Ying et al.

This work addresses the outlier removal problem in large-scale global structure-from-motion. In such applications, global outlier removal is very useful to mitigate the deterioration caused by mismatches in the feature point matching step. Unlike existing outlier removal methods, we exploit the structure in multiview geometry problems to propose a dimension reduced formulation, based on which two methods have been developed. The first method considers a convex relaxed $\ell_1$ minimization and is solved by a single linear programming (LP), whilst the second one approximately solves the ideal $\ell_0$ minimization by an iteratively reweighted method. The dimension reduction results in a significant speedup of the new algorithms. Further, the iteratively reweighted method can significantly reduce the possibility of removing true inliers. Realistic multiview reconstruction experiments demonstrated that, compared with state-of-the-art algorithms, the new algorithms are much more efficient and meanwhile can give improved solution. Matlab code for reproducing the results is available at \textit{https://github.com/FWen/OUTLR.git}.

Yang Su, Feng Yu, Yu Hu et al.

Autonomous drone racing in complex environments requires agile, high-speed flight while maintaining reliable obstacle avoidance. Differentiable-physics-based policy learning has recently demonstrated high sample efficiency and remarkable performance across various tasks, including agile drone flight and quadruped locomotion. However, applying such methods to drone racing remains difficult, as key objective like gate traversal are inherently hard to express as smooth, differentiable losses. To address these challenges, we propose DiffRacing, a novel vector field-augmented differentiable policy learning framework. DiffRacing integrates differentiable losses and vector fields into the training process to provide continuous and stable gradient signals, balancing obstacle avoidance and high-speed gate traversal. In addition, a differentiable Delta Action Model compensates for dynamics mismatch, enabling efficient sim-to-real transfer without explicit system identification. Extensive simulation and real-world experiments demonstrate that DiffRacing achieves superior sample efficiency, faster convergence, and robust flight performance, thereby demonstrating that vector fields can augment traditional gradient-based policy learning with a task-specific geometric prior.

Dexin Zuo, Ang Li, Wei Wang et al.

Object 6D pose estimation, a crucial task for robotics and augmented reality applications, becomes particularly challenging when dealing with novel objects whose 3D models are not readily available. To reduce dependency on 3D models, recent studies have explored one-reference-based pose estimation, which requires only a single reference view instead of a complete 3D model. However, existing methods that rely on real-valued coordinate regression suffer from limited global consistency due to the local nature of convolutional architectures and face challenges in symmetric or occluded scenarios owing to a lack of uncertainty modeling. We present CoordAR, a novel autoregressive framework for one-reference 6D pose estimation of unseen objects. CoordAR formulates 3D-3D correspondences between the reference and query views as a map of discrete tokens, which is obtained in an autoregressive and probabilistic manner. To enable accurate correspondence regression, CoordAR introduces 1) a novel coordinate map tokenization that enables probabilistic prediction over discretized 3D space; 2) a modality-decoupled encoding strategy that separately encodes RGB appearance and coordinate cues; and 3) an autoregressive transformer decoder conditioned on both position-aligned query features and the partially generated token sequence. With these novel mechanisms, CoordAR significantly outperforms existing methods on multiple benchmarks and demonstrates strong robustness to symmetry, occlusion, and other challenges in real-world tests.

Weixuan Tang, Danping Zou

We address the problem of estimating the poses of multiple instances of the source point cloud within a target point cloud. Existing solutions require sampling a lot of hypotheses to detect possible instances and reject the outliers, whose robustness and efficiency degrade notably when the number of instances and outliers increase. We propose to directly group the set of noisy correspondences into different clusters based on a distance invariance matrix. The instances and outliers are automatically identified through clustering. Our method is robust and fast. We evaluated our method on both synthetic and real-world datasets. The results show that our approach can correctly register up to 20 instances with an F1 score of 90.46% in the presence of 70% outliers, which performs significantly better and at least 10x faster than existing methods

Fanyi Xiao, Ling Pei, Lei Chu et al.

Sensor-based human activity recognition (HAR) is now a research hotspot in multiple application areas. With the rise of smart wearable devices equipped with inertial measurement units (IMUs), researchers begin to utilize IMU data for HAR. By employing machine learning algorithms, early IMU-based research for HAR can achieve accurate classification results on traditional classical HAR datasets, containing only simple and repetitive daily activities. However, these datasets rarely display a rich diversity of information in real-scene. In this paper, we propose a novel method based on deep learning for complex HAR in the real-scene. Specially, in the off-line training stage, the AMASS dataset, containing abundant human poses and virtual IMU data, is innovatively adopted for enhancing the variety and diversity. Moreover, a deep convolutional neural network with an unsupervised penalty is proposed to automatically extract the features of AMASS and improve the robustness. In the on-line testing stage, by leveraging advantages of the transfer learning, we obtain the final result by fine-tuning the partial neural network (optimizing the parameters in the fully-connected layers) using the real IMU data. The experimental results show that the proposed method can surprisingly converge in a few iterations and achieve an accuracy of 91.15% on a real IMU dataset, demonstrating the efficiency and effectiveness of the proposed method.

Boying Li, Danping Zou, Daniele Sartori et al.

We propose to integrate text objects in man-made scenes tightly into the visual SLAM pipeline. The key idea of our novel text-based visual SLAM is to treat each detected text as a planar feature which is rich of textures and semantic meanings. The text feature is compactly represented by three parameters and integrated into visual SLAM by adopting the illumination-invariant photometric error. We also describe important details involved in implementing a full pipeline of text-based visual SLAM. To our best knowledge, this is the first visual SLAM method tightly coupled with the text features. We tested our method in both indoor and outdoor environments. The results show that with text features, the visual SLAM system becomes more robust and produces much more accurate 3D text maps that could be useful for navigation and scene understanding in robotic or augmented reality applications.

Danping Zou, Yuanxin Wu, Ling Pei et al.

We propose a novel visual-inertial odometry approach that adopts structural regularity in man-made environments. Instead of using Manhattan world assumption, we use Atlanta world model to describe such regularity. An Atlanta world is a world that contains multiple local Manhattan worlds with different heading directions. Each local Manhattan world is detected on-the-fly, and their headings are gradually refined by the state estimator when new observations are coming. With fully exploration of structural lines that aligned with each local Manhattan worlds, our visual-inertial odometry method become more accurate and robust, as well as much more flexible to different kinds of complex man-made environments. Through extensive benchmark tests and real-world tests, the results show that the proposed approach outperforms existing visual-inertial systems in large-scale man-made environments

Rui Huang, Danping Zou, Richard Vaughan et al.

Image-based modeling techniques can now generate photo-realistic 3D models from images. But it is up to users to provide high quality images with good coverage and view overlap, which makes the data capturing process tedious and time consuming. We seek to automate data capturing for image-based modeling. The core of our system is an iterative linear method to solve the multi-view stereo (MVS) problem quickly and plan the Next-Best-View (NBV) effectively. Our fast MVS algorithm enables online model reconstruction and quality assessment to determine the NBVs on the fly. We test our system with a toy unmanned aerial vehicle (UAV) in simulated, indoor and outdoor experiments. Results show that our system improves the efficiency of data acquisition and ensures the completeness of the final model.

Yuanxin Wu, Danping Zou, Peilin Liu et al.

Magnetometer and inertial sensors are widely used for orientation estimation. Magnetometer usage is often troublesome, as it is prone to be interfered by onboard or ambient magnetic disturbance. The onboard soft-iron material distorts not only the magnetic field, but the magnetometer sensor frame coordinate and the cross-sensor misalignment relative to inertial sensors. It is desirable to conveniently put magnetic and inertial sensors information in a common frame. Existing methods either split the problem into successive intrinsic and cross-sensor calibrations, or rely on stationary accelerometer measurements which is infeasible in dynamic conditions. This paper formulates the magnetometer calibration and alignment to inertial sensors as a state estimation problem, and collectively solves the magnetometer intrinsic and cross-sensor calibrations, as well as the gyroscope bias estimation. Sufficient conditions are derived for the problem to be globally observable, even when no accelerometer information is used at all. An extended Kalman filter is designed to implement the state estimation and comprehensive test data results show the superior performance of the proposed approach. It is immune to acceleration disturbance and applicable potentially in any dynamic conditions.