Deep Rigid Instance Scene Flow

Ma, Wei-Chiu; Wang, Shenlong; Hu, Rui; Xiong, Yan; Urtasun, Raquel

doi:10.1109/cvpr.2019.00373

Cited by 131 publications

(107 citation statements)

References 49 publications

Supporting

Mentioning

107

Contrasting

Order By: Relevance

“…Moreover, the final covariance of X * k given Z, P k|N is given by P k|N = (I + P k J k ) −1 P k (27) We see the equations above allow performing optimal smoothing without involving at any time the matrix inverses Q −1 or P −1 k where P k denotes the forward covariance matrices, see (17). Indeed, in (22) and (23) each time those matrices are involved in an inversion operation there is a natural regularization term (I + •) involved as well.…”

Section: B the Backward Information Filter Forward Marginal (Bifm)mentioning

confidence: 99%

“…This is noticeable as there is a need to speed up computations for real-time applications, for which single precision is generally considered [45], [10], and most industrial aerospace computers still use single precision (32 bits or less) [17]. Besides, navigation methods relying on machine learning to improve some of their bricks [7], [6], or mimicking Kalman filters [19], are on the rise, and some recent methods even rely on Gauss-Newton as one of their bricks [27]. They also could benefit from single precision algorithms to speed up their training phase.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Factor Graph-Based Smoothing Without Matrix Inversion for Highly Precise Localization

Chauchat¹,

Barrau

Bonnabel

2021

IEEE Trans. Contr. Syst. Technol.

View full text Add to dashboard Cite

We consider the problem of localizing a manned, semi-autonomous, or autonomous vehicle in the environment using information coming from the vehicle's sensors, a problem known as navigation or simultaneous localization and mapping (SLAM) depending on the context. To infer knowledge from sensors' measurements, while drawing on a priori knowledge about the vehicle's dynamics, modern approaches solve an optimization problem to compute the most likely trajectory given all past observations, an approach known as smoothing. Improving smoothing solvers is an active field of research in the SLAM community. Most work is focused on reducing computation load by inverting the involved linear system while preserving its sparsity. The present paper raises an issue which, to the knowledge of the authors, has not been addressed yet: standard smoothing solvers require explicitly using the inverse of sensor noise covariance matrices. This means the parameters that reflect the noise magnitude must be sufficiently large for the smoother to properly function. When matrices are close to singular, which is the case when using high precision modern inertial measurement units (IMU), numerical issues necessarily arise, especially with 32-bits implementation demanded by most industrial aerospace applications. We discuss these issues and propose a solution that builds upon the Kalman filter to improve smoothing algorithms. We then leverage the results to devise a localization algorithm based on fusion of IMU and vision sensors. Successful real experiments using an actual car equipped with a tactical grade high performance IMU and a LiDAR illustrate the relevance of the approach to the field of autonomous vehicles.

show abstract

Section: B the Backward Information Filter Forward Marginal (Bifm)mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Factor Graph-Based Smoothing Without Matrix Inversion for Highly Precise Localization

Chauchat¹,

Barrau

Bonnabel

2021

IEEE Trans. Contr. Syst. Technol.

View full text Add to dashboard Cite

show abstract

“…A rigid plane model performs poorly when applied to deformable objects, and ego-motion estimation for highly dynamic scenes is hard. (Menze and Geiger, 2015) CSF (Lv et al, 2016) Deep Learning Fast Poor generalization PWOC-3D (Saxena et al, 2019) DRISF (Ma et al, 2019) Sparse-to-Dense Comparatively fast, good generalization Sensitive to distribution of matches SFF Schuster et al (2018c) SFF++ (ours)…”

Section: Related Workmentioning

confidence: 99%

“…Methods which are guided by semantic segmentation from deep neural networks will generalize badly to other domains, unless they are fine-tuned for the new task. Same is assumed for upcoming purely learning based approaches (Ma et al, 2019;Saxena et al, 2019) which are potentially even faster than our approach. SFF++ focuses especially on robustness across domains and applications.…”

Section: Related Workmentioning

confidence: 99%

SceneFlowFields++: Multi-frame Matching, Visibility Prediction, and Robust Interpolation for Scene Flow Estimation

et al. 2019

View full text Add to dashboard Cite

State-of-the-art scene flow algorithms pursue the conflicting targets of accuracy, run time, and robustness. With the successful concept of pixel-wise matching and sparse-to-dense interpolation, we shift the operating point in this field of conflicts towards universality and speed. Avoiding strong assumptions on the domain or the problem yields a more robust algorithm. This algorithm is fast because we avoid explicit regularization during matching, which allows an efficient computation. Using image information from multiple time steps and explicit visibility prediction based on previous results, we achieve competitive performances on different data sets. Our contributions and results are evaluated in comparative experiments. Overall, we present an accurate scene flow algorithm that is faster and more generic than any individual benchmark leader.

show abstract

“…DeepFlow interleaving between convolution and max-pooling layers in a multi-stage architecture. A recent work [21] adopted deep learning network and a strong prior in the motion field estimation. [21] has shown result outperform current state-of-art on the KITTI benchmark dataset [22].…”

Section: Convolutional Neural Network For Optical Flowmentioning

confidence: 99%

A Dense Optical Flow Field Estimation with Variational Refinement

Eng

Chang

Lim

et al. 2019

JETAP

View full text Add to dashboard Cite

Optical flow has long been a focus of research study in computer vision community. Researchers have established extensive work to solve the optical flow estimation. Among the published works, a notable work using variational energy minimization has beena baseline of optical flow estimation for a long time. Variational optical flow optimization solves an approximate global minimum in a well-defined non-linear Markov Energy formulation. It works by first linearizing the energy model and uses a numerical method specifically successive over-relaxation (SOR) method to solve the resulting linear model. An initialization scheme is required for optical flow field in this iterative optimization method. In the original work, a zero initialization is proposed and it works well on the various environments with photometric and geometric distortion. In this work, we have experimented with different flow field initialization scheme under various environment setting. We found out that variational refinement with a good initial flow estimate using state-of-art optical flow algorithms can further improve its accuracy performance.

show abstract

Deep Rigid Instance Scene Flow

Cited by 131 publications

References 49 publications

Factor Graph-Based Smoothing Without Matrix Inversion for Highly Precise Localization

Factor Graph-Based Smoothing Without Matrix Inversion for Highly Precise Localization

SceneFlowFields++: Multi-frame Matching, Visibility Prediction, and Robust Interpolation for Scene Flow Estimation

A Dense Optical Flow Field Estimation with Variational Refinement

Contact Info

Product

Resources

About