Bundle adjustment (BA) is a critical technique in various robotic applications such as simultaneous localization and mapping (SLAM), augmented reality (AR), and photogrammetry. BA optimizes parameters such as camera poses and 3D landmarks to align them with observations. With the growing importance of deep learning in perception systems, there is an increasing need to integrate BA with deep learning frameworks for enhanced reliability and performance. However, widely-used C++-based BA libraries, such as GTSAM, g$^2$o, and Ceres Solver, lack native integration with modern deep learning libraries like PyTorch. This limitation affects their flexibility, ease of debugging, and overall implementation efficiency. To address this gap, we introduce an eager-mode BA library seamlessly integrated with PyTorch with high efficiency. Our approach includes a sparsity-aware auto-differentiation design and GPU-accelerated sparse operations designed for 2nd-order optimization. Our eager-mode BA on GPU demonstrates substantial runtime efficiency, achieving an average speedup of 18.5$\times$, 22$\times$, and 23$\times$ across all benchmarks compared to GTSAM, g$^2$o, and Ceres, respectively.

Dense image correspondence is central to many applications, such as visual odometry, 3D reconstruction, object association, and re-identification. Historically, dense correspondence has been tackled separately for wide-baseline scenarios and optical flow estimation, despite the common goal of matching content between two images. In this paper, we develop a Unified Flow \& Matching model (UFM), which is trained on unified data for pixels that are co-visible in both source and target images. UFM uses a simple, generic transformer architecture that directly regresses the $(u,v)$ flow. It is easier to train and more accurate for large flows compared to the typical coarse-to-fine cost volumes in prior work. UFM is 28\% more accurate than state-of-the-art flow methods (Unimatch), while also having 62\% less error and 6.7x faster than dense wide-baseline matchers (RoMa). UFM is the first to demonstrate that unified training can outperform specialized approaches across both domains. This result enables fast, general-purpose correspondence and opens new directions for multi-modal, long-range, and real-time correspondence tasks.

Multi-view stereo omnidirectional distance estimation usually needs to build a cost volume with many hypothetical distance candidates. The cost volume building process is often computationally heavy considering the limited resources a mobile robot has. We propose a new geometry-informed way of distance candidates selection method which enables the use of a very small number of candidates and reduces the computational cost. We demonstrate the use of the geometry-informed candidates in a set of model variants. We find that by adjusting the candidates during robot deployment, our geometry-informed distance candidates also improve a pre-trained model’s accuracy if the extrinsics or the number of cameras changes. Without any re-training or fine-tuning, our models outperform models trained with evenly distributed distance candidates. Models are also released as hardware-accelerated versions with a new dedicated large-scale dataset. The project page, code, and dataset can be found at https://theairlab.org/gicandidates/.

Deep learning has had remarkable success in robotic perception, but its data-centric nature suffers when it comes to generalizing to ever-changing environments. By contrast, physics-based optimization generalizes better, but it does not perform as well in complicated tasks due to the lack of high-level semantic information and reliance on manual parametric tuning. To take advantage of these two complementary worlds, we present PyPose: a robotics-oriented, PyTorch-based library that combines deep perceptual models with physics-based optimization. PyPose's architecture is tidy and well-organized, it has an imperative style interface and is efficient and user-friendly, making it easy to integrate into real-world robotic applications. Besides, it supports parallel computing of any order gradients of Lie groups and Lie algebras and 2nd-order optimizers, such as trust region methods. Experiments show that PyPose achieves more than 10× speedup in computation compared to the state-of-the-art libraries. To boost future research, we provide concrete examples for several fields of robot learning, including SLAM, planning, control, and inertial navigation.

Detecting unseen instances based on multi-view templates is a challenging problem due to its open-world nature. Traditional methodologies, which primarily rely on $2 \mathrm{D}$ representations and matching techniques, are often inadequate in handling pose variations and occlusions. To solve this, we introduce VoxDet, a pioneer 3D geometry-aware framework that fully utilizes the strong 3D voxel representation and reliable voxel matching mechanism. VoxDet first ingeniously proposes template voxel aggregation (TVA) module, effectively transforming multi-view 2D images into 3D voxel features. By leveraging associated camera poses, these features are aggregated into a compact 3D template voxel. In novel instance detection, this voxel representation demonstrates heightened resilience to occlusion and pose variations. We also discover that a $3 \mathrm{D}$ reconstruction objective helps to pre-train the 2D-3D mapping in TVA. Second, to quickly align with the template voxel, VoxDet incorporates a Query Voxel Matching (QVM) module. The 2D queries are first converted into their voxel representation with the learned 2D-3D mapping. We find that since the 3D voxel representations encode the geometry, we can first estimate the relative rotation and then compare the aligned voxels, leading to improved accuracy and efficiency. In addition to method, we also introduce the first instance detection benchmark, RoboTools, where 20 unique instances are video-recorded with camera extrinsic. RoboTools also provides 24 challenging cluttered scenarios with more than $9 \mathrm{k}$ box annotations. Exhaustive experiments are conducted on the demanding LineMod-Occlusion, YCB-video, and RoboTools benchmarks, where VoxDet outperforms various $2 \mathrm{D}$ baselines remarkably with faster speed. To the best of our knowledge, VoxDet is the first to incorporate implicit 3D knowledge for 2D novel instance detection tasks.

In Simultaneous Localization And Mapping (SLAM) problems, high-level landmarks have the potential to build compact and informative maps compared to traditional point-based landmarks. In this work, we focus on the param-eterization of frequently used geometric primitives including points, lines, planes, ellipsoids, cylinders, and cones. We first present a unified representation based on quadrics, an algebraic representation of quadratic surfaces in 3D. Then we propose a decomposed model of quadrics that discloses the symmetry and degeneration properties of a primitive. Based on the decomposition, we develop geometrically meaningful quadrics factors for the graph-SLAM problem. Then in simulation, it is shown that the decomposed formulation has better efficiency and robustness to observation noises than baseline parame-terizations. Finally, in real-world experiments, the proposed back-end framework is demonstrated to be capable of building compact and regularized maps.

Stereo reconstruction models trained on small images do not generalize well to high-resolution data. Training a model on high-resolution image size faces difficulties of data availability and is often infeasible due to limited computing resources. In this work, we present the Occlusion-aware Recurrent binocular Stereo matching (ORStereo), which deals with these issues by only training on available low disparity range stereo images. ORStereo generalizes to unseen high-resolution images with large disparity ranges by formulating the task as residual updates and refinements of an initial prediction. ORStereo is trained on images with disparity ranges limited to 256 pixels, yet it can operate 4K-resolution input with over 1000 disparities using limited GPU memory. We test the model’s capability on both synthetic and real-world high-resolution images. Experimental results demonstrate that ORStereo achieves comparable performance on 4K-resolution images compared to state-of-the-art methods trained on large disparity ranges. Compared to the baseline methods that are only trained on low-resolution images, our method has 60% or less error on 4K-resolution images.

This work presents dense stereo reconstruction using high-resolution images for infrastructure inspections. The state-of-the-art stereo reconstruction methods, both learning and non-learning ones, consume too much computational resource on high-resolution data. Recent learning-based methods achieve top ranks on most benchmarks. However, they suffer from the generalization issue due to lack of task-specific training data. We propose to use a less resource demanding non-learning method, guided by a learning-based model, to handle high-resolution images and achieve accurate stereo reconstruction. The deep-learning model produces an initial disparity prediction with uncertainty for each pixel of the down-sampled stereo image pair. The uncertainty serves as a self-measurement of its generalization ability and the perpixel searching range around the initially predicted disparity. The downstream process performs a modified version of the Semi-Global Block Matching method with the up-sampled perpixel searching range. The proposed deep-learning assisted method is evaluated on the Middlebury dataset and high-resolution stereo images collected by our customized binocular stereo camera. The combination of learning and non-learning methods achieves better performance on 12 out of 15 cases of the Middlebury dataset. In our infrastructure inspection experiments, the average 3D reconstruction error is less than 0.004m.

Although Structure-from-Motion (SfM) as a maturing technique has been widely used in many applications, state-of-the-art SfM algorithms are still not robust enough in certain situations. For example, images for inspection purposes are often taken in close distance to obtain detailed textures, which will result in less overlap between images and thus decrease the accuracy of estimated motion. In this paper, we propose a LiDAR-enhanced SfM pipeline that jointly processes data from a rotating LiDAR and a stereo camera pair to estimate sensor motions. We show that incorporating LiDAR helps to effectively reject falsely matched images and significantly improve the model consistency in large-scale environments. Experiments are conducted in different environments to test the performance of the proposed pipeline and comparison results with the state-of-the-art SfM algorithms are reported.

We present a challenging dataset, the TartanAir, for robot navigation tasks and more. The data is collected in photo-realistic simulation environments with the presence of moving objects, changing light and various weather conditions. By collecting data in simulations, we are able to obtain multi-modal sensor data and precise ground truth labels such as the stereo RGB image, depth image, segmentation, optical flow, camera poses, and LiDAR point cloud. We set up large numbers of environments with various styles and scenes, covering challenging viewpoints and diverse motion patterns that are difficult to achieve by using physical data collection platforms. In order to enable data collection at such a large scale, we develop an automatic pipeline, including mapping, trajectory sampling, data processing, and data verification. We evaluate the impact of various factors on visual SLAM algorithms using our data. The results of state-of-the-art algorithms reveal that the visual SLAM problem is far from solved. Methods that show good performance on established datasets such as KITTI do not perform well in more difficult scenarios. Although we use the simulation, our goal is to push the limits of Visual SLAM algorithms in the real world by providing a challenging benchmark for testing new methods, while also using a large diverse training data for learning-based methods. Our dataset is available at http://theairlab.org/tartanair-dataset.