Query Point Research Articles

P2M: A Fast Solver for Querying Distance from Point to Mesh Surface

Most of the existing point-to-mesh distance query solvers, such as Proximity Query Package (PQP), Embree and Fast Closest Point Query (FCPW), are based on bounding volume hierarchy (BVH). The hierarchical organizational structure enables one to eliminate the vast majority of triangles that do not help find the closest point. In this paper, we develop a totally different algorithmic paradigm, named P2M , to speed up point-to-mesh distance queries. Our original intention is to precompute a KD tree (KDT) of mesh vertices to approximately encode the geometry of a mesh surface containing vertices, edges and faces. However, it is very likely that the closest primitive to the query point is an edge e (resp., a face f ), but the KDT reports a mesh vertex υ instead. We call υ an interceptor of e (resp., f ). The main contribution of this paper is to invent a simple yet effective interception inspection rule and an efficient flooding interception inspection algorithm for quickly finding out all the interception pairs. Once the KDT and the interception table are precomputed, the query stage proceeds by first searching the KDT and then looking up the interception table to retrieve the closest geometric primitive. Statistics show that our query algorithm runs many times faster than the state-of-the-art solvers.

GPU-Based Algorithms for Processing the k Nearest-Neighbor Query on Spatial Data Using Partitioning and Concurrent Kernel Execution

Algorithms for answering the k nearest-neighbor (k-NN) query are widely used for queries in spatial databases and for distance classification of a group of query points against a reference dataset to derive the dominating feature class. GPU devices have significantly more processing cores than CPUs and faster device memory than the main memory accessed by CPUs, thus, providing higher computing power for processing demanding queries like the k-NN. However, since device and/or main memory may not be able to host an entire, rather big, reference and query datasets, storing these datasets in a fast secondary device, like a solid state disk (SSD), and partially retrieve the required, at each stage, partitions is, in many practical cases, a feasible solution. We propose and implement the first GPU-based algorithms for processing the k-NN query for big reference and query spatial data stored on SSDs. Based on 3d synthetic and real big spatial data, we experimentally compare these algorithms and highlight the most efficient algorithmic variation. This variation utilizes a CUDA feature known as Concurrent Kernel Execution, to further improve its performance.

Robust table structure recognition with dynamic queries enhanced detection transformer

We present a new table structure recognition (TSR) approach, called TSRFormer, to robustly recognize the structures of complex tables with geometrical distortions from various table images. Unlike previous methods, we formulate table separation line prediction as a line regression problem instead of an image segmentation problem and propose a new two-stage dynamic queries enhanced DETR based separation line regression approach, named DQ-DETR, to predict separation lines from table images directly. Compared to Vallina DETR, we propose three improvements in DQ-DETR to make the two-stage DETR framework work efficiently and effectively for the separation line prediction task: 1) A new query design, named Dynamic Query, to decouple single line query into separable point queries which could intuitively improve the localization accuracy for regression tasks; 2) A dynamic queries based progressive line regression approach to progressively regressing points on the line which further enhances localization accuracy for distorted tables; 3) A prior-enhanced matching strategy to solve the slow convergence issue of DETR. After separation line prediction, a simple relation network based cell merging module is used to recover spanning cells. With these new techniques, our TSRFormer achieves state-of-the-art performance on several benchmark datasets, including SciTSR, PubTabNet, WTW, FinTabNet, and cTDaR TrackB2-Modern. Furthermore, we have validated the robustness and high localization accuracy of our approach to tables with complex structures, borderless cells, large blank spaces, empty or spanning cells as well as distorted or even curved shapes on a more challenging real-world in-house dataset.

IAN: Instance-Augmented Net for 3D Instance Segmentation

When aggregating local information from neighbors, prevailing 3D instance segmentation backbones only leverage 3D coordinates to find neighboring points without identifying whether these points are from the same object as the query point, which causes the model to gather excessive noisy features. Besides, traditional backbones fail to fully utilize multi-resolution information. Therefore, previous methods have difficulty in segmenting targets in cluttered scenes. To tackle these issues, we propose Instance-Augmented Net (IAN). The keys to our approach are Instance-Augmented Block (IAB), Instance-Augmented Upsampler (IAU), and Attentive Fusion (AF). In IAB, for each foreground point, we leverage its instance information to filter out noisy neighbors from other objects. We also propose IAU to apply this instance-augmented strategy to the upsampling process. Furthermore, to retain comprehensive information, we upsample multi-resolution feature maps and adopt attention generated by AF to fuse them. Notably, by encoding neighborhood information, AF can generate attention at point-level adaptively. Moreover, to further test the generality of models, we present Clutter and Occlusion (CAO), a new 3D instance segmentation dataset tailored for robotic grasping tasks. Extensive experiments on S3DIS, ScanNet and CAO show the effectiveness of our IAN.

R 3 : Record-Replay-Retroaction for Database-Backed Applications

Developers would benefit greatly from time travel: being able to faithfully replay past executions and retroactively execute modified code on past events. Currently, replay and retroaction are impractical because they require expensively capturing fine-grained timing information to reproduce concurrent accesses to shared state. In this paper, we propose practical time travel for database-backed applications , an important class of distributed applications that access shared state through transactions. We present R 3 , a novel Record-Replay-Retroaction tool. R 3 implements a lightweight interceptor to record concurrency information for applications at transaction-level granularity, enabling replay and retroaction with minimal overhead. We address key challenges in both replay and retroaction. First, we design a novel algorithm for faithfully reproducing application requests running with snapshot isolation, allowing R 3 to support most production DBMSs. Second, we develop a retroactive execution mechanism that provides high fidelity with the original trace while supporting nearly arbitrary code modifications. We demonstrate how R 3 simplifies debugging for real, hard-to-reproduce concurrency bugs from popular open-source web applications. We evaluate R 3 using TPC-C and microservice workloads and show that R 3 always-on recording has a small performance overhead (<25% for point queries but <0.1% for complex transactions like in TPC-C) during normal application execution and that R 3 can retroactively execute bugfixed code over recorded traces within 0.11--0.78× of the original execution time.

DPText-DETR: Towards Better Scene Text Detection with Dynamic Points in Transformer

Recently, Transformer-based methods, which predict polygon points or Bezier curve control points for localizing texts, are popular in scene text detection. However, these methods built upon detection transformer framework might achieve sub-optimal training efficiency and performance due to coarse positional query modeling. In addition, the point label form exploited in previous works implies the reading order of humans, which impedes the detection robustness from our observation. To address these challenges, this paper proposes a concise Dynamic Point Text DEtection TRansformer network, termed DPText-DETR. In detail, DPText-DETR directly leverages explicit point coordinates to generate position queries and dynamically updates them in a progressive way. Moreover, to improve the spatial inductive bias of non-local self-attention in Transformer, we present an Enhanced Factorized Self-Attention module which provides point queries within each instance with circular shape guidance. Furthermore, we design a simple yet effective positional label form to tackle the side effect of the previous form. To further evaluate the impact of different label forms on the detection robustness in real-world scenario, we establish an Inverse-Text test set containing 500 manually labeled images. Extensive experiments prove the high training efficiency, robustness, and state-of-the-art performance of our method on popular benchmarks. The code and the Inverse-Text test set are available at https://github.com/ymy-k/DPText-DETR.

Few-Shot 3D Point Cloud Semantic Segmentation via Stratified Class-Specific Attention Based Transformer Network

3D point cloud semantic segmentation aims to group all points into different semantic categories, which benefits important applications such as point cloud scene reconstruction and understanding. Existing supervised point cloud semantic segmentation methods usually require large-scale annotated point clouds for training and cannot handle new categories. While a few-shot learning method was proposed recently to address these two problems, it suffers from high computational complexity caused by graph construction and inability to learn fine-grained relationships among points due to the use of pooling operations. In this paper, we further address these problems by developing a new multi-layer transformer network for few-shot point cloud semantic segmentation. In the proposed network, the query point cloud features are aggregated based on the class-specific support features in different scales. Without using pooling operations, our method makes full use of all pixel-level features from the support samples. By better leveraging the support features for few-shot learning, the proposed method achieves the new state-of-the-art performance, with 15% less inference time, over existing few-shot 3D point cloud segmentation models on the S3DIS dataset and the ScanNet dataset. Our code is available at https://github.com/czzhang179/SCAT.

SEFormer: Structure Embedding Transformer for 3D Object Detection

Effectively preserving and encoding structure features from objects in irregular and sparse LiDAR points is a crucial challenge to 3D object detection on the point cloud. Recently, Transformer has demonstrated promising performance on many 2D and even 3D vision tasks. Compared with the fixed and rigid convolution kernels, the self-attention mechanism in Transformer can adaptively exclude the unrelated or noisy points and is thus suitable for preserving the local spatial structure in the irregular LiDAR point cloud. However, Transformer only performs a simple sum on the point features, based on the self-attention mechanism, and all the points share the same transformation for value. A such isotropic operation cannot capture the direction-distance-oriented local structure, which is essential for 3D object detection. In this work, we propose a Structure-Embedding transFormer (SEFormer), which can not only preserve the local structure as a traditional Transformer but also have the ability to encode the local structure. Compared to the self-attention mechanism in traditional Transformer, SEFormer learns different feature transformations for value points based on the relative directions and distances to the query point. Then we propose a SEFormer-based network for high-performance 3D object detection. Extensive experiments show that the proposed architecture can achieve SOTA results on the Waymo Open Dataset, one of the most significant 3D detection benchmarks for autonomous driving. Specifically, SEFormer achieves 79.02% mAP, which is 1.2% higher than existing works. https://github.com/tdzdog/SEFormer.

Relationships between tail entropies and local intrinsic dimensionality and their use for estimation and feature representation

The local intrinsic dimensionality (LID) model assesses the complexity of data within the vicinity of a query point, through the growth rate of the probability measure within an expanding neighborhood. In this paper, we show how LID is asymptotically related to the entropy of the lower tail of the distribution of distances from the query. We establish relationships for cumulative Shannon entropy, entropy power, Bregman formulation of cumulative Kullback–Leibler divergence, and generalized Tsallis entropy variants. Leveraging these relationships, we propose four new estimators of LID, one of them expressible in an intuitive analytic form. We investigate the effectiveness of these new estimators, as well as the effectiveness of entropy power as the basis for feature representations in classification.

Forward inclusion functions for ray-tracing implicit surfaces

Using Lipschitz bounds as linear inclusion functions, we show that both Lipschitz-based ray-tracing and bottom-up inclusion functions can be used together in the same framework. We propose asymmetrical forward inclusion functions that are exact at the query point and can better encode the function’s behavior on a given interval; therefore well suited for iterative processing. We show how to derive the linear and quadratic versions of these inclusion functions either by bounding the derivatives or building bottom-up inclusion functions and combining these two. We show our results on density fields defined from point primitives with compactly supported kernels and Gaussian kernels as well as Hermite radial basis functions. We demonstrate that our method provides noticeable improvement for grazing rays and transparent rendering.

EAR-Oracle: On Efficient Indexing for Distance Queries between Arbitrary Points on Terrain Surface

Due to the advancement of geo-positioning technology, the terrain data has become increasingly popular and has drawn a lot of research effort from both academia and industry. The distance computation on the terrain surface is a fundamental and important problem that is widely applied in geographical information systems and 3D modeling. As could be observed from the existing studies, online computation of the distance on the terrain surface is very expensive. All existing index-based methods are only efficient under the case where the distance query must be performed among a small set of predefined points-of-interest known apriori. But, in general cases, they could not scale up to sizable datasets due to their intolerable oracle building time and space consumption. In this paper, we studied the arbitrary point-to-arbitrary point distance query on the terrain surface in which no assumption is imposed on the query points, and the distance query could be performed between any two arbitrary points. We propose an indexing structure, namely Efficient Arbitrary Point-to-Arbitrary Point Distance Oracle (EAR-Oracle), with theoretical guarantee on the accuracy, oracle building time, oracle size and query time. Our experiments demonstrate that our oracle enjoys excellent scalability and it scales up to enormous terrain surfaces but none of the existing index-based methods could be able to. Besides, it significantly outperforms all existing online computation methods by orders of magnitude in terms of the query time.

A stable and effective calibration method for defocused cameras using synthetic speckle patterns

Well-focused and sharp target images are essential in traditional calibration, but they generally cause difficulty and inconvenience in image acquisition in practical applications. To address this problem, a stable and effective calibration method for defocused cameras is proposed in this study. Firstly, two speckle patterns are created, one of which has control points to estimate the initial value while the other, used for point pair match registration, is without control points. They are subsequently displayed on a fixed monitor and captured at each camera pose. Secondly, the initial guess for the control points is estimated using the proposed method. The retrieval of query points can then be done with high accuracy using the digital image correlation algorithm. Finally, all the calibration parameters, including lens distortion, are calculated using the plane calibration algorithm. The experiment in this study demonstrates the practicality and effectiveness of the proposed method. The accuracy and precision were quantitatively evaluated by the mean reprojection error, which was 0.057 pixels.

2D k-th nearest neighbour statistics: a highly informative probe of galaxy clustering

ABSTRACT Beyond standard summary statistics are necessary to summarize the rich information on non-linear scales in the era of precision galaxy clustering measurements. For the first time, we introduce the 2D k-th nearest neighbour (kNN) statistics as a summary statistic for discrete galaxy fields. This is a direct generalization of the standard 1D kNN by disentangling the projected galaxy distribution from the redshift-space distortion signature along the line-of-sight. We further introduce two different flavours of 2D kNNs that trace different aspects of the galaxy field: the standard flavour which tabulates the distances between galaxies and random query points, and a ‘DD’ flavour that tabulates the distances between galaxies and galaxies. We showcase the 2D kNNs’ strong constraining power both through theoretical arguments and by testing on realistic galaxy mocks. Theoretically, we show that 2D kNNs are computationally efficient and directly generate other statistics such as the popular two-point correlation function (2PCF), voids probability function, and counts-in-cell statistics. In a more practical test, we apply the 2D kNN statistics to simulated galaxy mocks that fold in a large range of observational realism and recover parameters of the underlying extended halo occupation distribution (HOD) model that includes velocity bias and galaxy assembly bias. We find unbiased and significantly tighter constraints on all aspects of the HOD model with the 2D kNNs, both compared to the standard 1D kNN, and the classical redshift-space 2PCF.

SLBRIN: A Spatial Learned Index Based on BRIN

The spatial learned index constructs a spatial index by learning the spatial distribution, which performs a lower cost of storage and query than the spatial indices. The current update strategies of spatial learned indices can only solve limited updates at the cost of query performance. We propose a novel spatial learned index structure based on a Block Range Index (SLBRIN for short). Its core idea is to cooperate history range and current range to satisfy a fast spatial query and efficient index update simultaneously. SLBRIN deconstructs the update transaction into three parallel operations and optimizes them based on the temporal proximity of spatial distribution. SLBRIN also provides the spatial query strategy with the spatial learned index and spatial location code, including point query, range query and kNN query. Experiments on synthetic and real datasets demonstrate that SLBRIN clearly outperforms traditional spatial indices and state-of-the-art spatial learned indices in the cost of storage and query. Moreover, in the simulated real-time update scenario, SLBRIN has the faster and more stable query performance while satisfying efficient updates.

Spatial Autocorrelation Incorporated Machine Learning Model for Geotechnical Subsurface Modeling

Machine learning models for spatial prediction have been applied in various types of research. However, spatial relation has not been fully considered in modeling, since the Cartesian coordinates of the observed points are directly employed as the location information for machine learning features. This study presents a machine learning modeling process which incorporates spatial autocorrelation for geotechnical subsurface modeling. A new set of features called the Euclidean distance field (EDF) was generated based on the distance between the query points and the observed boreholes in order to incorporate spatial autocorrelation into the machine learning model. Principal component analysis (PCA) was performed to reduce the increasing dimensionality of the dataset caused by the EDF features. Optimized machine learning models based on several popular algorithms (Support Vector Machine, Gaussian Process Regression, Artificial Neural Network, and k-Nearest Neighbor) were employed for predicting several geotechnical information as the targets. The results showed that the optimized machine learning models constructed with the EDF modeling approach generate a slightly lower Root Mean Square Error (RMSE) score compared to the model with the direct XY coordinate approach by 0.041, 0.046, 1.302, and 1.561 for ground surface elevation, groundwater level, SPT-N value, and percent finer than 0.075 mm sieve, respectively. Both modeling approaches performed well for USCS-based soil classification with the EDF model having slightly improved classification accuracy by 0.72%. Furthermore, the model can perform balance multiclass classification as indicated by the >95% precision, recall, f1-score, and balanced accuracy score. These results indicate that spatial autocorrelation has a noticeable effect. Hence, it needs to be considered to improve the overall performance of spatial machine learning modeling. Comparison of geotechnical subsurface predictions generated based on different machine learning algorithms showed that the selection of the best-performing model based only on the lowest prediction error is not appropriate for spatial prediction modeling. Therefore, thorough analysis of the predicted data by visualization is necessary in the selection process for spatial prediction modeling.

Efficient simulation of non-stationary non-homogeneous wind field: Fusion of multi-dimensional interpolation and NUFFT

The stochastic wave approach is a new realization of the spectral representation method and has been widely used for the simulation of stationary non-homogeneous or non-stationary homogeneous wind fields. However, it inevitably creates many unnecessary simulation points due to the invoking of the Fast Fourier Transform (FFT) and suffers from low efficiency in non-stationary non-homogeneous scenarios due to the incapability to accelerate the summation of trigonometric functions. In this study, an efficient approach is developed for the simulation of non-stationary non-homogeneous wind field. Central to this approach is a multi-dimensional interpolation used to decouple the evolutionary spectra accompanied by the application of non-uniform FFT (NUFFT) to expedite the simulation. A surrogate model that accurately approximates the 3D evolutionary spectrum related to height, frequency, and time is established by the 3D reduced Hermite interpolation, which enables a successful separation of the frequency components. Then, the NUFFT is applied to speed up the summation of trigonometric functions over wavenumbers with a well-designed non-uniformly distributed sampling points, which effectively reduces the segments used in the Fourier transform. Since there is no required mapping between the wavenumbers and the locations of simulation points in NUFFT, a significant number of unnecessary simulation points in the FFT scenario is avoided. Moreover, the wind samples at different locations can be obtained by directly altering the query points of NUFFT, thus the efficiency of the enhanced approach is almost independent of the number of simulation points. The numerical example in simulating the non-stationary non-homogeneous wind field of a long-span cable-stayed bridge demonstrates the efficiency of the developed approach and validates the effectiveness of the simulated wind samples in view of evolutionary spectrum and coherence function.

Learning Continuous-Time Dynamics With Attention.

Learning the hidden dynamics from sequence data is crucial. Attention mechanism can be introduced to spotlight on the region of interest for sequential learning. Traditional attention was measured between a query and a sequence based on a discrete-time state trajectory. Such a mechanism could not characterize the irregularly-sampled sequence data. This paper presents an attentive differential network (ADN) where the attention over continuous-time dynamics is developed. The continuous-time attention is performed over the dynamics at all time. The missing information in irregular or sparse samples can be seamlessly compensated and attended. Self attention is computed to find the attended state trajectory. However, the memory cost for attention score between a query and a sequence is demanding since self attention treats all time instants as query points in an ordinary differential equation solver. This issue is tackled by imposing the causality constraint in causal ADN (CADN) where the query is merged up to current time. To enhance the model robustness, this study further explores a latent CADN where the attended dynamics are calculated in an encoder-decoder structure via Bayesian learning. Experiments on the irregularly-sampled actions, dialogues and bio-signals illustrate the merits of the proposed methods in action recognition, emotion recognition and mortality prediction, respectively.

A Category-Contrastive Guided-Graph Convolutional Network Approach for the Semantic Segmentation of Point Clouds

The semantic segmentation of light detection and ranging (LiDAR) point clouds plays an important role in 3D scene intelligent perception and semantic modeling. The unstructured, sparse and uneven characteristics of point clouds pose great challenges to the representation of the local geometric shapes, which degrades semantic segmentation performance. To address the challenges of describing local geometric shapes due to unstructured and sparse 3D point clouds, this paper proposes a category-contrastive-guided graph convolutional network (CGGC-Net) for the semantic segmentation of LiDAR point clouds. First, a detailed geometric structure of the raw point clouds is encoded to represent the inherent geometric pattern within the local neighborhood. At the same time, the geometric structures information is transmitted across multiple layers, so that the geometric structure encoding information containing different receptive fields and richer neighborhood spatial structure can be aggregated. Following this, the graph convolution neural network uses the edge convolution layer to adaptively describe the semantic correlation between the query point and its neighboring points, and combines the attention mechanism to gather the surrounding feature information to the query point. As a result, the graph convolution neural network and attention mechanism are iteratively stacked for the aggregation and fusion of spatial context semantic information, to generate highly discriminative semantic feature representation. Finally, the super parameters of the model are learned through a multitask optimization strategy guided by category-aware contrastive loss and cross-entropy loss. Experiments are conducted on the public SemanticKITTI dataset and the Stanford Large-Scale 3D Indoor Spaces dataset to demonstrate the effectiveness and reliability of the proposed CGGC-Net from both quantitative and qualitative perspectives. The results indicate its capability of automatically classifying LiDAR point clouds, with a mean intersection-over-union (mIoU) of 58.4%. Moreover, multiple comparative experiments also demonstrate the superior performance of the proposed method, exceeding state-of-the-art methods.

Prototype Adaption and Projection for Few- and Zero-Shot 3D Point Cloud Semantic Segmentation.

In this work, we address the challenging task of few-shot and zero-shot 3D point cloud semantic segmentation. The success of few-shot semantic segmentation in 2D computer vision is mainly driven by the pre-training on large-scale datasets like imagenet. The feature extractor pre-trained on large-scale 2D datasets greatly helps the 2D few-shot learning. However, the development of 3D deep learning is hindered by the limited volume and instance modality of datasets due to the significant cost of 3D data collection and annotation. This results in less representative features and large intra-class feature variation for few-shot 3D point cloud segmentation. As a consequence, directly extending existing popular prototypical methods of 2D few-shot classification/segmentation into 3D point cloud segmentation won't work as well as in 2D domain. To address this issue, we propose a Query-Guided Prototype Adaption (QGPA) module to adapt the prototype from support point clouds feature space to query point clouds feature space. With such prototype adaption, we greatly alleviate the issue of large feature intra-class variation in point cloud and significantly improve the performance of few-shot 3D segmentation. Besides, to enhance the representation of prototypes, we introduce a Self-Reconstruction (SR) module that enables prototype to reconstruct the support mask as well as possible. Moreover, we further consider zero-shot 3D point cloud semantic segmentation where there is no support sample. To this end, we introduce category words as semantic information and propose a semantic-visual projection model to bridge the semantic and visual spaces. Our proposed method surpasses state-of-the-art algorithms by a considerable 7.90% and 14.82% under the 2-way 1-shot setting on S3DIS and ScanNet benchmarks, respectively.

Gradient-Enhanced Bayesian Optimization via Acquisition Ensembles with Application to Reinforcement Learning

Bayesian optimization (BO) has shown great promise as a data-efficient strategy for the global optimization of expensive, black-box functions in a plethora of control applications. Traditional BO is derivative-free, as it solely relies on observations of a performance function to find its optimum. Recently, so-called first-order BO methods have been proposed that additionally exploit gradient information of the performance function to accelerate convergence. First-order BO methods mostly utilize standard acquisition functions, while indirectly using gradient information in the kernel structure to learn more accurate probabilistic surrogates for the performance function. In this work, we present a gradient-enhanced BO method that directly exploits performance function (zeroth-order) and its corresponding gradient (first-order) evaluations in the acquisition function. To this end, a novel gradient-based acquisition function is proposed that can identify stationary points of the performance optimization problem. We then leverage ideas from multi-objective optimization to develop an effective strategy for finding query points that optimally tradeoff between a zeroth-order acquisition function and the proposed gradient-based acquisition function. We show how the proposed acquisition-ensemble gradient-enhanced BO (AEGEBO) method enables accelerating convergence of policy-based reinforcement learning by combining noisy observations of the reward function and its gradient that can be directly estimated from closed-loop data. The performance of AEGBO is compared to standard BO and the well-known REINFORCE algorithm on a benchmark LQR problem, for which we consistently observe significantly improved performance over a limited data budget.