arXiv subjects

Pei Sun

Publications and source records attributed to Pei Sun.

50 records · Page 3Linked to original sources

arXiv2021

Offboard 3D Object Detection from Point Cloud Sequences

While current 3D object recognition research mostly focuses on the real-time, onboard scenario, there are many offboard use cases of perception that are largely under-explored, such as using machines to automatically generate high-quality 3D labels. Existing 3D object detectors fail to satisfy the high-quality requirement for offboard uses due to the limited input and speed constraints. In this paper, we propose a novel offboard 3D object detection pipeline using point cloud sequence data. Observing that different frames capture complementary views of objects, we design the offboard detector to make use of the temporal points through both multi-frame object detection and novel object-centric refinement models. Evaluated on the Waymo Open Dataset, our pipeline named 3D Auto Labeling shows significant gains compared to the state-of-the-art onboard detectors and our offboard baselines. Its performance is even on par with human labels verified through a human label study. Further experiments demonstrate the application of auto labels for semi-supervised learning and provide extensive analysis to validate various design choices.

cs.CV↗

arXiv2021

Exact solutions of the $C_n$ quantum spin chain

We study the exact solutions of quantum integrable model associated with the $C_n$ Lie algebra, with either a periodic or an open one with off-diagonal boundary reflections, by generalizing the nested off-diagonal Bethe ansatz method. Taking the $C_3$ as an example we demonstrate how the generalized method works. We give the fusion structures of the model and provide a way to close fusion processes. Based on the resulted operator product identities among fused transfer matrices and some necessary additional constraints such as asymptotic behaviors and relations at some special points, we obtain the eigenvalues of transfer matrices and parameterize them as homogeneous $T-Q$ relations in the periodic case or inhomogeneous ones in the open case. We also give the exact solutions of the $C_n$ model with an off-diagonal open boundary condition. The method and results in this paper can be generalized to other high rank integrable models associated with other Lie algebras.

math-ph↗

arXiv2021

Range Conditioned Dilated Convolutions for Scale Invariant 3D Object Detection

This paper presents a novel 3D object detection framework that processes LiDAR data directly on its native representation: range images. Benefiting from the compactness of range images, 2D convolutions can efficiently process dense LiDAR data of a scene. To overcome scale sensitivity in this perspective view, a novel range-conditioned dilation (RCD) layer is proposed to dynamically adjust a continuous dilation rate as a function of the measured range. Furthermore, localized soft range gating combined with a 3D box-refinement stage improves robustness in occluded areas, and produces overall more accurate bounding box predictions. On the public large-scale Waymo Open Dataset, our method sets a new baseline for range-based 3D detection, outperforming multiview and voxel-based methods over all ranges with unparalleled performance at long range detection.

cs.CV↗

arXiv2020

Exact ground state and elementary excitations of a topological spin chain

A novel Bethe Ansatz scheme is proposed to calculate physical properties of quantum integrable systems without $U(1)$ symmetry. As an example, the anti-periodic XXZ spin chain, a typical correlated many-body system embedded in a topological manifold, is examined. Conserved "momentum" and "charge" operators are constructed despite the absence of translational invariance and $U(1)$ symmetry. The ground state energy and elementary excitations are derived exactly. It is found that two intrinsic fractional (one half) zero modes accounting for the double degeneracy exist in the eigenstates. The elementary excitations show quite a different picture from that of a periodic chain. This method can be applied to other quantum integrable models either with or without $U(1)$ symmetry.

math-ph↗

arXiv2020

SurfelGAN: Synthesizing Realistic Sensor Data for Autonomous Driving

Autonomous driving system development is critically dependent on the ability to replay complex and diverse traffic scenarios in simulation. In such scenarios, the ability to accurately simulate the vehicle sensors such as cameras, lidar or radar is essential. However, current sensor simulators leverage gaming engines such as Unreal or Unity, requiring manual creation of environments, objects and material properties. Such approaches have limited scalability and fail to produce realistic approximations of camera, lidar, and radar data without significant additional work. In this paper, we present a simple yet effective approach to generate realistic scenario sensor data, based only on a limited amount of lidar and camera data collected by an autonomous vehicle. Our approach uses texture-mapped surfels to efficiently reconstruct the scene from an initial vehicle pass or set of passes, preserving rich information about object 3D geometry and appearance, as well as the scene conditions. We then leverage a SurfelGAN network to reconstruct realistic camera images for novel positions and orientations of the self-driving vehicle and moving objects in the scene. We demonstrate our approach on the Waymo Open Dataset and show that it can synthesize realistic camera data for simulated scenarios. We also create a novel dataset that contains cases in which two self-driving vehicles observe the same scene at the same time. We use this dataset to provide additional evaluation and demonstrate the usefulness of our SurfelGAN model.

cs.CV↗

arXiv2020

Scalability in Perception for Autonomous Driving: Waymo Open Dataset

The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited in the scale and variation of the environments they capture, even though generalization within and between operating regions is crucial to the overall viability of the technology. In an effort to help align the research community's contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset. Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies. It is 15x more diverse than the largest camera+LiDAR dataset available based on our proposed diversity metric. We exhaustively annotated this data with 2D (camera image) and 3D (LiDAR) bounding boxes, with consistent identifiers across frames. Finally, we provide strong baselines for 2D as well as 3D detection and tracking tasks. We further study the effects of dataset size and generalization across geographies on 3D detection methods. Find data, code and more up-to-date information at http://www.waymo.com/open.

cs.CV↗

arXiv2019

StarNet: Targeted Computation for Object Detection in Point Clouds

Detecting objects from LiDAR point clouds is an important component of self-driving car technology as LiDAR provides high resolution spatial information. Previous work on point-cloud 3D object detection has re-purposed convolutional approaches from traditional camera imagery. In this work, we present an object detection system called StarNet designed specifically to take advantage of the sparse and 3D nature of point cloud data. StarNet is entirely point-based, uses no global information, has data dependent anchors, and uses sampling instead of learned region proposals. We demonstrate how this design leads to competitive or superior performance on the large Waymo Open Dataset and the KITTI detection dataset, as compared to convolutional baselines. In particular, we show how our detector can outperform a competitive baseline on Pedestrian detection on the Waymo Open Dataset by more than 7 absolute mAP while being more computationally efficient. We show how our redesign---namely using only local information and using sampling instead of learned proposals---leads to a significantly more flexible and adaptable system: we demonstrate how we can vary the computational cost of a single trained StarNet without retraining, and how we can target proposals towards areas of interest with priors and heuristics. Finally, we show how our design allows for incorporating temporal context by using detections from previous frames to target computation of the detector, which leads to further improvements in performance without additional computational cost.

cs.CV↗

arXiv2019

End-to-End Multi-View Fusion for 3D Object Detection in LiDAR Point Clouds

Recent work on 3D object detection advocates point cloud voxelization in birds-eye view, where objects preserve their physical dimensions and are naturally separable. When represented in this view, however, point clouds are sparse and have highly variable point density, which may cause detectors difficulties in detecting distant or small objects (pedestrians, traffic signs, etc.). On the other hand, perspective view provides dense observations, which could allow more favorable feature encoding for such cases. In this paper, we aim to synergize the birds-eye view and the perspective view and propose a novel end-to-end multi-view fusion (MVF) algorithm, which can effectively learn to utilize the complementary information from both. Specifically, we introduce dynamic voxelization, which has four merits compared to existing voxelization methods, i) removing the need of pre-allocating a tensor with fixed size; ii) overcoming the information loss due to stochastic point/voxel dropout; iii) yielding deterministic voxel embeddings and more stable detection outcomes; iv) establishing the bi-directional relationship between points and voxels, which potentially lays a natural foundation for cross-view feature fusion. By employing dynamic voxelization, the proposed feature fusion architecture enables each point to learn to fuse context information from different views. MVF operates on points and can be naturally extended to other approaches using LiDAR point clouds. We evaluate our MVF model extensively on the newly released Waymo Open Dataset and on the KITTI dataset and demonstrate that it significantly improves detection accuracy over the comparable single-view PointPillars baseline.

cs.CV↗

arXiv2019

Off-diagonal Bethe Ansatz on the $so(5)$ spin chain

The $so(5)$ (i.e., $B_2$) quantum integrable spin chains with both periodic and non-diagonal boundaries are studied via the off-diagonal Bethe Ansatz method. By using the fusion technique, sufficient operator product identities (comparing to those in [1]) to determine the spectrum of the transfer matrices are derived. For the periodic case, we recover the results obtained in \cite{NYReshetikhin1}, while for the non-diagonal boundary case, a new inhomogeneous $T-Q$ relation is constructed. The present method can be directly generalized to deal with the $so(2n+1)$ (i.e., $B_n$) quantum integrable spin chains with general boundaries.

math-ph↗

arXiv2019

Exact solution of an integrable anisotropic $J_1-J_2$ spin chain model

An integrable anisotropic Heisenberg spin chain with nearest-neighbour couplings, next-nearest-neighbour couplings and scalar chirality terms is constructed. After proving the integrability, we obtain the exact solution of the system. The ground state and the elementary excitations are also studied. It is shown that the spinon excitation of the present model possesses a novel triple arched structure. The elementary excitation is gapless if the anisotropic parameter $η$ is real while the elementary excitation has an enhanced gap by the next-nearest-neighbour and chiral three-spin interactions if the anisotropic parameter $η$ is imaginary. The method of this paper provides a general way to construct new integrable models with next-nearest-neighbour interactions.

math-ph↗

arXiv2019

Surface energy and elementary excitations of the XXZ spin chain with arbitrary boundary fields

The thermodynamic properties of the XXZ spin chain with integrable open boundary conditions at the gaped region (i.e., the anisotropic parameter $η$ being a real number) are investigated.It is shown that the contribution of the inhomogeneous term in the $T-Q$ relation of the ground state and elementary excited state can be neglected when the size of the system $N$ tends to infinity. The surface energy and elementary excitations induced by the unparallel boundary magnetic fields are obtained.

cond-mat.str-el↗

arXiv2018

Bethe states of the trigonometric SU(3) spin chain with generic open boundaries

By combining the algebraic Bethe ansatz and the off-diagonal Bethe ansatz, we investigate the trigonometric SU(3) model with generic open boundaries. The eigenvalues of the transfer matrix are given in terms of an inhomogeneous T-Q relation, and the corresponding eigenstates are expressed in terms of nested Bethe-type eigenstates which have well-defined homogeneous limit. This exact solution provides a basis for further analyzing the thermodynamic properties and correlation functions of the anisotropic models associated with higher rank algebras.

math-ph↗

arXiv2017

On the Bethe states of the one-dimensional supersymmetric t-J model with generic open boundaries

By combining the algebraic Bethe ansatz and the off-diagonal Bethe ansatz, we investigate the supersymmetric t-J model with generic open boundaries. The eigenvalues of the transfer matrix are given in terms of an inhomogeneous T-Q relation, and the corresponding eigenstates are expressed in terms of nested Bethe states which have well-defined homogeneous limit. This exact solution provides basis for further analyzing the thermodynamic properties and correlation functions of the model.

math-ph↗

arXiv2009

Efficient Signal-Time Coding Design and its Application in Wireless Gaussian Relay Networks

Signal-time coding, which combines the traditional encoding/modulation mode in the signal domain with signal pulse phase modulation in the time domain, was proposed to improve the information flow rate in relay networks. In this paper, we mainly focus on the efficient signal-time coding design. We first derive an explicit iterative algorithm to estimate the maximum number of available codes given the code length of signal-time coding, and then present an iterative construction method of codebooks. It is shown that compared with conventional computer search, the proposed iterative construction method can reduce the complexity greatly. Numerical results will also indicate that the new constructed codebook is optimal in terms of coding rate. To minimize the buffer size needed to store the codebook while keeping a relatively high efficiency, we shall propose a combinatorial construction method. We will then consider applications in wireless Gaussian relay networks. It will be shown that in the three node network model, the mixed transmission by using two-hop and direct transmissions is not always a good option.

cs.IT↗