Depth and ego-motion estimations are essential for the localization and navigation of autonomous robots and autonomous driving. Recent studies make it possible to learn the per-pixel depth and ego-motion from the unlabeled monocular video. A novel unsupervised training framework is proposed with 3D hierarchical refinement and augmentation using explicit 3D geometry. In this framework, the depth and pose estimations are hierarchically and mutually coupled to refine the estimated pose layer by layer. The intermediate view image is proposed and synthesized by warping the pixels in an image with the estimated depth and coarse pose. Then, the residual pose transformation can be estimated from the new view image and the image of the adjacent frame to refine the coarse pose. The iterative refinement is implemented in a differentiable manner in this paper, making the whole framework optimized uniformly. Meanwhile, a new image augmentation method is proposed for the pose estimation by synthesizing a new view image, which creatively augments the pose in 3D space but gets a new augmented 2D image. The experiments on KITTI demonstrate that our depth estimation achieves state-of-the-art performance and even surpasses recent approaches that utilize other auxiliary tasks. Our visual odometry outperforms all recent unsupervised monocular learning-based methods and achieves competitive performance to the geometry-based method, ORB-SLAM2 with back-end optimization.
Recent visual odometry (VO) methods incorporating geometric algorithm into deep-learning architecture have shown outstanding performance on the challenging monocular VO task. Despite encouraging results are shown, previous methods ignore the requirement of generalization capability under noisy environment and various scenes. To address this challenging issue, this work first proposes a novel optical flow network (PANet). Compared with previous methods that predict optical flow as a direct regression task, our PANet computes optical flow by predicting it into the discrete position space with optical flow probability volume, and then converting it to optical flow. Next, we improve the bundle adjustment module to fit the self-supervised training pipeline by introducing multiple sampling, ego-motion initialization, dynamic damping factor adjustment, and Jacobi matrix weighting. In addition, a novel normalized photometric loss function is advanced to improve the depth estimation accuracy. The experiments show that the proposed system not only achieves comparable performance with other state-of-the-art self-supervised learning-based methods on the KITTI dataset, but also significantly improves the generalization capability compared with geometry-based, learning-based and hybrid VO systems on the noisy KITTI and the challenging outdoor (KAIST) scenes.
We propose MC-CIM, a compute-in-memory (CIM) framework for robust, yet low power, Bayesian edge intelligence. Deep neural networks (DNN) with deterministic weights cannot express their prediction uncertainties, thereby pose critical risks for applications where the consequences of mispredictions are fatal such as surgical robotics. To address this limitation, Bayesian inference of a DNN has gained attention. Using Bayesian inference, not only the prediction itself, but the prediction confidence can also be extracted for planning risk-aware actions. However, Bayesian inference of a DNN is computationally expensive, ill-suited for real-time and/or edge deployment. An approximation to Bayesian DNN using Monte Carlo Dropout (MC-Dropout) has shown high robustness along with low computational complexity. Enhancing the computational efficiency of the method, we discuss a novel CIM module that can perform in-memory probabilistic dropout in addition to in-memory weight-input scalar product to support the method. We also propose a compute-reuse reformulation of MC-Dropout where each successive instance can utilize the product-sum computations from the previous iteration. Even more, we discuss how the random instances can be optimally ordered to minimize the overall MC-Dropout workload by exploiting combinatorial optimization methods. Application of the proposed CIM-based MC-Dropout execution is discussed for MNIST character recognition and visual odometry (VO) of autonomous drones. The framework reliably gives prediction confidence amidst non-idealities imposed by MC-CIM to a good extent. Proposed MC-CIM with 16x31 SRAM array, 0.85 V supply, 16nm low-standby power (LSTP) technology consumes 27.8 pJ for 30 MC-Dropout instances of probabilistic inference in its most optimal computing and peripheral configuration, saving 43% energy compared to typical execution.
In this paper, we present TANDEM a real-time monocular tracking and dense mapping framework. For pose estimation, TANDEM performs photometric bundle adjustment based on a sliding window of keyframes. To increase the robustness, we propose a novel tracking front-end that performs dense direct image alignment using depth maps rendered from a global model that is built incrementally from dense depth predictions. To predict the dense depth maps, we propose Cascade View-Aggregation MVSNet (CVA-MVSNet) that utilizes the entire active keyframe window by hierarchically constructing 3D cost volumes with adaptive view aggregation to balance the different stereo baselines between the keyframes. Finally, the predicted depth maps are fused into a consistent global map represented as a truncated signed distance function (TSDF) voxel grid. Our experimental results show that TANDEM outperforms other state-of-the-art traditional and learning-based monocular visual odometry (VO) methods in terms of camera tracking. Moreover, TANDEM shows state-of-the-art real-time 3D reconstruction performance.
Visual odometry algorithms tend to degrade when facing low-textured scenes -from e.g. human-made environments-, where it is often difficult to find a sufficient number of point features. Alternative geometrical visual cues, such as lines, which can often be found within these scenarios, can become particularly useful. Moreover, these scenarios typically present structural regularities, such as parallelism or orthogonality, and hold the Manhattan World assumption. Under these premises, in this work, we introduce MSC-VO, an RGB-D -based visual odometry approach that combines both point and line features and leverages, if exist, those structural regularities and the Manhattan axes of the scene. Within our approach, these structural constraints are initially used to estimate accurately the 3D position of the extracted lines. These constraints are also combined next with the estimated Manhattan axes and the reprojection errors of points and lines to refine the camera pose by means of local map optimization. Such a combination enables our approach to operate even in the absence of the aforementioned constraints, allowing the method to work for a wider variety of scenarios. Furthermore, we propose a novel multi-view Manhattan axes estimation procedure that mainly relies on line features. MSC-VO is assessed using several public datasets, outperforming other state-of-the-art solutions, and comparing favourably even with some SLAM methods.
In this paper, we propose a robust edge-direct visual odometry (VO) based on CNN edge detection and Shi-Tomasi corner optimization. Four layers of pyramids were extracted from the image in the proposed method to reduce the motion error between frames. This solution used CNN edge detection and Shi-Tomasi corner optimization to extract information from the image. Then, the pose estimation is performed using the Levenberg-Marquardt (LM) algorithm and updating the keyframes. Our method was compared with the dense direct method, the improved direct method of Canny edge detection, and ORB-SLAM2 system on the RGB-D TUM benchmark. The experimental results indicate that our method achieves better robustness and accuracy.
Estimating the motion of the camera together with the 3D structure of the scene from a monocular vision system is a complex task that often relies on the so-called scene rigidity assumption. When observing a dynamic environment, this assumption is violated which leads to an ambiguity between the ego-motion of the camera and the motion of the objects. To solve this problem, we present a self-supervised learning framework for 3D object motion field estimation from monocular videos. Our contributions are two-fold. First, we propose a two-stage projection pipeline to explicitly disentangle the camera ego-motion and the object motions with dynamics attention module, called DAM. Specifically, we design an integrated motion model that estimates the motion of the camera and object in the first and second warping stages, respectively, controlled by the attention module through a shared motion encoder. Second, we propose an object motion field estimation through contrastive sample consensus, called CSAC, taking advantage of weak semantic prior (bounding box from an object detector) and geometric constraints (each object respects the rigid body motion model). Experiments on KITTI, Cityscapes, and Waymo Open Dataset demonstrate the relevance of our approach and show that our method outperforms state-of-the-art algorithms for the tasks of self-supervised monocular depth estimation, object motion segmentation, monocular scene flow estimation, and visual odometry.
Lines provide the significantly richer geometric structural information about the environment than points, so lines are widely used in recent Visual Odometry (VO) works. Since VO with lines use line tracking results to locate and map, line tracking is a crucial component in VO. Although the state-of-the-art line tracking methods have made great progress, they are still heavily dependent on line detection or the predicted line segments. In order to relieve the dependencies described above to track line segments completely, accurately, and robustly at higher computational efficiency, we propose a structure-aware Line tracking algorithm based entirely on Optical Flow (LOF). Firstly, we propose a gradient-based strategy to sample pixels on lines that are suitable for line optical flow calculation. Then, in order to align the lines by fully using the structural relationship between the sampled points on it and effectively removing the influence of sampled points on it occluded by other objects, we propose a two-step structure-aware line segment alignment method. Furthermore, we propose a line refinement method to refine the orientation, position, and endpoints of the aligned line segments. Extensive experimental results demonstrate that the proposed LOF outperforms the state-of-the-art performance in line tracking accuracy, robustness, and efficiency, which also improves the location accuracy and robustness of VO system with lines.
We present a novel self-supervised algorithm named MotionHint for monocular visual odometry (VO) that takes motion constraints into account. A key aspect of our approach is to use an appropriate motion model that can help existing self-supervised monocular VO (SSM-VO) algorithms to overcome issues related to the local minima within their self-supervised loss functions. The motion model is expressed with a neural network named PPnet. It is trained to coarsely predict the next pose of the camera and the uncertainty of this prediction. Our self-supervised approach combines the original loss and the motion loss, which is the weighted difference between the prediction and the generated ego-motion. Taking two existing SSM-VO systems as our baseline, we evaluate our MotionHint algorithm on the standard KITTI benchmark. Experimental results show that our MotionHint algorithm can be easily applied to existing open-sourced state-of-the-art SSM-VO systems to greatly improve the performance by reducing the resulting ATE by up to 28.73%.
In the future, extraterrestrial expeditions will not only be conducted by rovers but also by flying robots. The technical demonstration drone Ingenuity, that just landed on Mars, will mark the beginning of a new era of exploration unhindered by terrain traversability. Robust self-localization is crucial for that. Cameras that are lightweight, cheap and information-rich sensors are already used to estimate the ego-motion of vehicles. However, methods proven to work in man-made environments cannot simply be deployed on other planets. The highly repetitive textures present in the wastelands of Mars pose a huge challenge to descriptor matching based approaches. In this paper, we present an advanced robust monocular odometry algorithm that uses efficient optical flow tracking to obtain feature correspondences between images and a refined keyframe selection criterion. In contrast to most other approaches, our framework can also handle rotation-only motions that are particularly challenging for monocular odometry systems. Furthermore, we present a novel approach to estimate the current risk of scale drift based on a principal component analysis of the relative translation information matrix. This way we obtain an implicit measure of uncertainty. We evaluate the validity of our approach on all sequences of a challenging real-world dataset captured in a Mars-like environment and show that it outperforms state-of-the-art approaches.
Over the last decade, one of the most relevant public datasets for evaluating odometry accuracy is the KITTI dataset. Beside the quality and rich sensor setup, its success is also due to the online evaluation tool, which enables researchers to benchmark and compare algorithms. The results are evaluated on the test subset solely, without any knowledge about the ground truth, yielding unbiased, overfit free and therefore relevant validation for robot localization based on cameras, 3D laser or combination of both. However, as any sensor setup, it requires prior calibration and rectified stereo images are provided, introducing dependence on the default calibration parameters. Given that, a natural question arises if a better set of calibration parameters can be found that would yield higher odometry accuracy. In this paper, we propose a new approach for one shot calibration of the KITTI dataset multiple camera setup. The approach yields better calibration parameters, both in the sense of lower calibration reprojection errors and lower visual odometry error. We conducted experiments where we show for three different odometry algorithms, namely SOFT2, ORB-SLAM2 and VISO2, that odometry accuracy is significantly improved with the proposed calibration parameters. Moreover, our odometry, SOFT2, in conjunction with the proposed calibration method achieved the highest accuracy on the official KITTI scoreboard with 0.53% translational and 0.0009 deg/m rotational error, outperforming even 3D laser-based methods.
It is fundamental for personal robots to reliably navigate to a specified goal. To study this task, PointGoal navigation has been introduced in simulated Embodied AI environments. Recent advances solve this PointGoal navigation task with near-perfect accuracy (99.6% success) in photo-realistically simulated environments, assuming noiseless egocentric vision, noiseless actuation, and most importantly, perfect localization. However, under realistic noise models for visual sensors and actuation, and without access to a "GPS and Compass sensor," the 99.6%-success agents for PointGoal navigation only succeed with 0.3%. In this work, we demonstrate the surprising effectiveness of visual odometry for the task of PointGoal navigation in this realistic setting, i.e., with realistic noise models for perception and actuation and without access to GPS and Compass sensors. We show that integrating visual odometry techniques into navigation policies improves the state-of-the-art on the popular Habitat PointNav benchmark by a large margin, improving success from 64.5% to 71.7% while executing 6.4 times faster.
Meshes are commonly used as 3D maps since they encode the topology of the scene while being lightweight. Unfortunately, 3D meshes are mathematically difficult to handle directly because of their combinatorial and discrete nature. Therefore, most approaches generate 3D meshes of a scene after fusing depth data using volumetric or other representations. Nevertheless, volumetric fusion remains computationally expensive both in terms of speed and memory. In this paper, we leapfrog these intermediate representations and build a 3D mesh directly from a depth map and the sparse landmarks triangulated with visual odometry. To this end, we formulate a non-smooth convex optimization problem that we solve using a primal-dual method. Our approach generates a smooth and accurate 3D mesh that substantially improves the state-of-the-art on direct mesh reconstruction while running in real-time.
Feature tracking is the building block of many applications such as visual odometry, augmented reality, and target tracking. Unfortunately, the state-of-the-art vision-based tracking algorithms fail in surgical images due to the challenges imposed by the nature of such environments. In this paper, we proposed a novel and unified deep learning-based approach that can learn how to track features reliably as well as learn how to detect such reliable features for tracking purposes. The proposed network dubbed as Deep-PT, consists of a tracker network which is a convolutional neural network simulating cross-correlation in terms of deep learning and two fully connected networks that operate on the output of intermediate layers of the tracker to detect features and predict trackability of the detected points. The ability to detect features based on the capabilities of the tracker distinguishes the proposed method from previous algorithms used in this area and improves the robustness of the algorithms against dynamics of the scene. The network is trained using multiple datasets due to the lack of specialized dataset for feature tracking datasets and extensive comparisons are conducted to compare the accuracy of Deep-PT against recent pixel tracking algorithms. As the experiments suggest, the proposed deep architecture deliberately learns what to track and how to track and outperforms the state-of-the-art methods.
Mobile AR applications benefit from fast initialization to display world-locked effects instantly. However, standard visual odometry or SLAM algorithms require motion parallax to initialize (see Figure 1) and, therefore, suffer from delayed initialization. In this paper, we present a 6-DoF monocular visual odometry that initializes instantly and without motion parallax. Our main contribution is a pose estimator that decouples estimating the 5-DoF relative rotation and translation direction from the 1-DoF translation magnitude. While scale is not observable in a monocular vision-only setting, it is still paramount to estimate a consistent scale over the whole trajectory (even if not physically accurate) to avoid AR effects moving erroneously along depth. In our approach, we leverage the fact that depth errors are not perceivable to the user during rotation-only motion. However, as the user starts translating the device, depth becomes perceivable and so does the capability to estimate consistent scale. Our proposed algorithm naturally transitions between these two modes. We perform extensive validations of our contributions with both a publicly available dataset and synthetic data. We show that the proposed pose estimator outperforms the classical approaches for 6-DoF pose estimation used in the literature in low-parallax configurations. We release a dataset for the relative pose problem using real data to facilitate the comparison with future solutions for the relative pose problem. Our solution is either used as a full odometry or as a preSLAM component of any supported SLAM system (ARKit, ARCore) in world-locked AR effects on platforms such as Instagram and Facebook.
Event-based cameras are biologically inspired sensors that output events, i.e., asynchronous pixel-wise brightness changes in the scene. Their high dynamic range and temporal resolution of a microsecond makes them more reliable than standard cameras in environments of challenging illumination and in high-speed scenarios, thus developing odometry algorithms based solely on event cameras offers exciting new possibilities for autonomous systems and robots. In this paper, we propose a novel stereo visual odometry method for event cameras based on feature detection and matching with careful feature management, while pose estimation is done by reprojection error minimization. We evaluate the performance of the proposed method on two publicly available datasets: MVSEC sequences captured by an indoor flying drone and DSEC outdoor driving sequences. MVSEC offers accurate ground truth from motion capture, while for DSEC, which does not offer ground truth, in order to obtain a reference trajectory on the standard camera frames we used our SOFT visual odometry, one of the highest ranking algorithms on the KITTI scoreboards. We compared our method to the ESVO method, which is the first and still the only stereo event odometry method, showing on par performance on the MVSEC sequences, while on the DSEC dataset ESVO, unlike our method, was unable to handle outdoor driving scenario with default parameters. Furthermore, two important advantages of our method over ESVO are that it adapts tracking frequency to the asynchronous event rate and does not require initialization.
Perspective-n-Point-and-Line (P$n$PL) algorithms aim at fast, accurate, and robust camera localization with respect to a 3D model from 2D-3D feature correspondences, being a major part of modern robotic and AR/VR systems. Current point-based pose estimation methods use only 2D feature detection uncertainties, and the line-based methods do not take uncertainties into account. In our setup, both 3D coordinates and 2D projections of the features are considered uncertain. We propose PnP(L) solvers based on EPnP and DLS for the uncertainty-aware pose estimation. We also modify motion-only bundle adjustment to take 3D uncertainties into account. We perform exhaustive synthetic and real experiments on two different visual odometry datasets. The new PnP(L) methods outperform the state-of-the-art on real data in isolation, showing an increase in mean translation accuracy by 18% on a representative subset of KITTI, while the new uncertain refinement improves pose accuracy for most of the solvers, e.g. decreasing mean translation error for the EPnP by 16% compared to the standard refinement on the same dataset. The code is available at https://alexandervakhitov.github.io/uncertain-pnp/.
We present a new solution to tracking and mapping with an event camera. The motion of the camera contains both rotation and translation, and the displacements happen in an arbitrarily structured environment. As a result, the image matching may no longer be represented by a low-dimensional homographic warping, thus complicating an application of the commonly used Image of Warped Events (IWE). We introduce a new solution to this problem by performing contrast maximization in 3D. The 3D location of the rays cast for each event is smoothly varied as a function of a continuous-time motion parametrization, and the optimal parameters are found by maximizing the contrast in a volumetric ray density field. Our method thus performs joint optimization over motion and structure. The practical validity of our approach is supported by an application to AGV motion estimation and 3D reconstruction with a single vehicle-mounted event camera. The method approaches the performance obtained with regular cameras, and eventually outperforms in challenging visual conditions.
Building vehicles capable of operating without human supervision requires the determination of the agent's pose. Visual Odometry (VO) algorithms estimate the egomotion using only visual changes from the input images. The most recent VO methods implement deep-learning techniques using convolutional neural networks (CNN) extensively, which add a substantial cost when dealing with high-resolution images. Furthermore, in VO tasks, more input data does not mean a better prediction; on the contrary, the architecture may filter out useless information. Therefore, the implementation of computationally efficient and lightweight architectures is essential. In this work, we propose the RAM-VO, an extension of the Recurrent Attention Model (RAM) for visual odometry tasks. RAM-VO improves the visual and temporal representation of information and implements the Proximal Policy Optimization (PPO) algorithm to learn robust policies. The results indicate that RAM-VO can perform regressions with six degrees of freedom from monocular input images using approximately 3 million parameters. In addition, experiments on the KITTI dataset demonstrate that RAM-VO achieves competitive results using only 5.7% of the available visual information.
The incremental poses computed through odometry can be integrated over time to calculate the pose of a device with respect to an initial location. The resulting global pose may be used to formulate a second, consistency based, loss term in a deep odometry setting. In such cases where multiple losses are imposed on a network, the uncertainty over each output can be derived to weigh the different loss terms in a maximum likelihood setting. However, when imposing a constraint on the integrated transformation, due to how only odometry is estimated at each iteration of the algorithm, there is no information about the uncertainty associated with the global pose to weigh the global loss term. In this paper, we associate uncertainties with the output poses of a deep odometry network and propagate the uncertainties through each iteration. Our goal is to use the estimated covariance matrix at each incremental step to weigh the loss at the corresponding step while weighting the global loss term using the compounded uncertainty. This formulation provides an adaptive method to weigh the incremental and integrated loss terms against each other, noting the increase in uncertainty as new estimates arrive. We provide quantitative and qualitative analysis of pose estimates and show that our method surpasses the accuracy of the state-of-the-art Visual Odometry approaches. Then, uncertainty estimates are evaluated and comparisons against fixed baselines are provided. Finally, the uncertainty values are used in a realistic example to show the effectiveness of uncertainty quantification for localization.