Cloud Object Research Articles

SkyPIE: A Fast & Accurate Oracle for Object Placement

Cloud object stores offer vastly different price points for object storage as a function of workload and geography. Poor object placement can thus lead to significant cost overheads. Prior cost-saving techniques attempt to optimize placement policies on the fly, deciding object placements for each object individually. In practice, these techniques do not scale to the size of the modern cloud. In this work, we leverage the static nature and pay-per-use pricing model of cloud environments to explore a different approach. Rather than computing object placements on the fly, we precompute a SkyPIE oracle---a lookup structure representing all possible placement policies and the workloads for which they are optimal. Internally, SkyPIE represents placement policies as a matrix of cost-hyperplanes, which we effectively precompute through pruning and convex optimization. By leveraging a fast geometric algorithm, online queries then are 1 to 8 orders of magnitude faster but as accurate as Integer-Linear-Programming. This makes exact optimization tractable for real workloads and we show >10x cost savings compared to state-of-the-art heuristic approaches.

3D Point Cloud Object Detection Method Based on Multi-Scale Dynamic Sparse Voxelization.

Perception plays a crucial role in ensuring the safety and reliability of autonomous driving systems. However, the recognition and localization of small objects in complex scenarios still pose challenges. In this paper, we propose a point cloud object detection method based on dynamic sparse voxelization to enhance the detection performance of small objects. This method employs a specialized point cloud encoding network to learn and generate pseudo-images from point cloud features. The feature extraction part uses sliding windows and transformer-based methods. Furthermore, multi-scale feature fusion is performed to enhance the granularity of small object information. In this experiment, the term "small object" refers to objects such as cyclists and pedestrians, which have fewer pixels compared to vehicles with more pixels, as well as objects of poorer quality in terms of detection. The experimental results demonstrate that, compared to the PointPillars algorithm and other related algorithms on the KITTI public dataset, the proposed algorithm exhibits improved detection accuracy for cyclist and pedestrian target objects. In particular, there is notable improvement in the detection accuracy of objects in the moderate and hard quality categories, with an overall average increase in accuracy of about 5%.

Public cloud object storage auditing: Design, implementation, and analysis

Cloud storage auditing is a technique that enables a user to remotely check the integrity of the outsourced data in the cloud storage. Although researchers have proposed various protocols for cloud storage auditing, the proposed schemes are theoretical in nature, which are not fit for existing mainstream cloud storage service practices. To bridge this gap, this paper proposes a cloud storage auditing system that works for current mainstream cloud object storage services. We design the proposed system over existing proof of data possession (PDP) schemes and make them practical as well as usable in the real world. Specifically, we propose an architecture that separates the compute and storage functionalities of a storage auditing scheme. Because cloud object storage only provides read and write interfaces, we leverage a cloud virtual machine to implement the user-defined computations that are needed in a PDP scheme. We store the authentication tags of the outsourced data as an independent object to allow existing popular cloud storage applications, e.g., file online previewing. We also present a cost model to analyze the economic cost of a cloud storage auditing scheme. The cost model allows a user to balance security, efficiency, and economic cost by tuning various system parameters. We implemented, open-sourced the proposed system over a mainstream cloud object storage service. Experimental analysis shows that the proposed system is pretty efficient and promising for a production environment usage. Specifically, for a 40 GB sized data, the proposed system only incurs 1.66% additional storage cost, 3796 bytes communication cost, 2.9 seconds maximum auditing time cost, and 0.9 CNY per auditing monetary cost.

Regional Cloud Forecast Verification Using Standard, Spatial, and Object-Oriented Methods

Abstract This study assesses the accuracy of the Coupled Ocean–Atmosphere Mesoscale Prediction System (COAMPS) forecasts for clouds within stable and unstable environments (thereafter refers as “stable” and “unstable” clouds). This evaluation is conducted by comparing these forecasts against satellite retrievals through a combination of traditional, spatial, and object-based methods. To facilitate this assessment, the Model Evaluation Tools (MET) community tool is employed. The findings underscore the significance of fine-tuning the MET parameters to achieve a more accurate representation of the features under scrutiny. The study’s results reveal that when employing traditional pointwise statistics (e.g., frequency bias and equitable threat score), there is consistency in the results whether calculated from Method for Object-Based Diagnostic Evaluation (MODE)-based objects or derived from the complete fields. Furthermore, the object-based statistics offer valuable insights, indicating that COAMPS generally predicts cloud object locations accurately, though the spread of these predicted locations tends to increase with time. It tends to overpredict the object area for unstable clouds while underpredicting it for stable clouds over time. These results are in alignment with the traditional pointwise bias scores for the entire grid. Overall, the spatial metrics provided by the object-based verification methods emerge as crucial and practical tools for the validation of cloud forecasts. Significance Statement As the general Navy meteorological and oceanographic (METOC) community engages in collaboration with the broader scientific community, our goal is to harness community tools like MET for the systematic evaluation of weather forecasts, with a specific focus on variables crucial to the Navy. Clouds, given their significant impact on visibility, hold particular importance in our investigations. Cloud forecasts pose unique challenges, primarily attributable to the intricate physics governing cloud development and the complexity of representing these processes within numerical models. Cloud observations are also constrained by limitations, arising from both top-down satellite measurements and bottom-up ground-based measurements. This study illustrates that, with a comprehensive understanding of community tools, cloud forecasts can be consistently verified. This verification encompasses traditional evaluation methods, measuring general qualities such as bias and root-mean-squared error, as well as newer techniques like spatial and object-based methods designed to account for displacement errors.

AFRNet: Anchor-Free Object Detection Using Roadside LiDAR in Urban Scenes

In urban settings, roadside infrastructure LiDAR is a ground-based remote sensing system that collects 3D sparse point clouds for the traffic object detection of vehicles, pedestrians, and cyclists. Current anchor-free algorithms for 3D point cloud object detection based on roadside infrastructure face challenges related to inadequate feature extraction, disregard for spatial information in large 3D scenes, and inaccurate object detection. In this study, we propose AFRNet, a two-stage anchor-free detection network, to address the aforementioned challenges. We propose a 3D feature extraction backbone based on the large sparse kernel convolution (LSKC) feature set abstraction module, and incorporate the CBAM attention mechanism to enhance the large scene feature extraction capability and the representation of the point cloud features, enabling the network to prioritize the object of interest. After completing the first stage of center-based prediction, we propose a refinement method based on attentional feature fusion, where fused features incorporating raw point cloud features, voxel features, BEV features, and key point features are used for the second stage of refinement to complete the detection of 3D objects. To evaluate the performance of our detection algorithms, we conducted experiments using roadside LiDAR data from the urban traffic dataset DAIR-V2X, based on the Beijing High-Level Automated Driving Demonstration Area. The experimental results show that AFRNet has an average of 5.27 percent higher detection accuracy than CenterPoint for traffic objects. Comparative tests further confirm that our method achieves high accuracy in roadside LiDAR object detection.

Equal Emphasis on Data and Network: A Two-Stage 3D Point Cloud Object Detection Algorithm with Feature Alignment

Three-dimensional object detection is a pivotal research topic in computer vision, aiming to identify and locate objects in three-dimensional space. It has wide applications in various fields such as geoscience, autonomous driving, and drone navigation. The rapid development of deep learning techniques has led to significant advancements in 3D object detection. However, with the increasing complexity of applications, 3D object detection faces a series of challenges such as data imbalance and the effectiveness of network models. Specifically, in an experiment, our investigation revealed a notable discrepancy in the LiDAR reflection intensity within a point cloud scene, with stronger intensities observed in proximity and weaker intensities observed at a distance. Furthermore, we have also noted a substantial disparity in the number of foreground points compared to the number of background points. Especially in 3D object detection, the foreground point is more important than the background point, but it is usually downsampled without discrimination in the subsequent processing. With the objective of tackling these challenges, we work from both data and network perspectives, designing a feature alignment filtering algorithm and a two-stage 3D object detection network. Firstly, in order to achieve feature alignment, we introduce a correction equation to decouple the relationship between distance and intensity and eliminate the attenuation effect of intensity caused by distance. Then, a background point filtering algorithm is designed by using the aligned data to alleviate the problem of data imbalance. At the same time, we take into consideration the fact that the accuracy of semantic segmentation plays a crucial role in 3D object detection. Therefore, we propose a two-stage deep learning network that integrates spatial and spectral information, in which a feature fusion branch is designed and embedded in the semantic segmentation backbone. Through a series of experiments on the KITTI dataset, it is proven that the proposed method achieves the following average precision (AP_R40) values for easy, moderate, and hard difficulties, respectively: car (Iou 0.7)—89.23%, 80.14%, and 77.89%; pedestrian (Iou 0.5)—52.32%, 45.47%, and 38.78%; and cyclist (Iou 0.5)—76.41%, 61.92%, and 56.39%. By emphasizing both data quality optimization and efficient network architecture, the performance of the proposed method is made comparable to other state-of-the-art methods.

High-speed imaging system to detect stellar occultations by Kuiper belt and Oort cloud objects

High-speed imaging system to detect stellar occultations by Kuiper belt and Oort cloud objects

Methods for Validating HRRR Simulated Cloud Properties for Different Weather Phenomena Using Satellite and Radar Observations

Abstract In this study, we evaluate the ability of the High-Resolution Rapid Refresh (HRRR) model to forecast cloud characteristics through comparison of observed and simulated satellite brightness temperatures (BTs) and radar reflectivity during different weather phenomena in December 2021: the Mayfield, Kentucky, tornado on 11 December, a heavy snow event in Minnesota from 10 to 11 December, and the Midwest derecho on 15 December. This is done to illustrate the importance of examining model accuracy across a range of weather phenomena. Observation and forecast objects were created using the Method for Object-Based Diagnostic Evaluation (MODE). HRRR accurately depicted the spatial displacements between observation cloud (defined using BTs) and radar reflectivity objects, namely, the centers of cloud objects are to the east of the radar objects for the tornado and derecho events, and generally west of the radar objects for the snow event. However, HRRR had higher (less intense) simulated BTs and higher (more intense) radar reflectivity than the observations for the tornado event. Simulated radar reflectivity is higher and BTs are lower than the observations during the middle of the snow event. Also, simulated radar reflectivity is higher and BTs are lower than the observations during the derecho event. Of the three weather events, the HRRR forecasts are most accurate for the snow event, based on the object-based threat score, followed by the derecho and tornado events. The tornado event has lower accuracy because matches between paired simulated and observation objects are worse than for the snow event, with less similarity in size forecast objects and greater distance between paired object centers. Significance Statement The purpose of this study is to assess the accuracy of forecast cloud and radar objects, defined using simulated satellite brightness temperatures and radar reflectivity, from the High-Resolution Rapid Refresh (HRRR) model. This assessment was conducted for a tornado, snow, and derecho event from December 2021. Results from these three events indicate that the HRRR model accurately represents the observed displacement between the center of cloud and radar objects for the tornado and derecho events, and is the most accurate overall for the snow event.

DGC-TnT: Enhancing Point Cloud Object Classification by Dynamic Graph Convolutions with Transformer in Transformer

DGC-TnT: Enhancing Point Cloud Object Classification by Dynamic Graph Convolutions with Transformer in Transformer

Planes geometry verification program written in Python

In recent years, there has been significant progress in sensor and scanning technologies in the computer vision and 3D visualization field. This has led to an increased interest in processing and analysing point clouds, which are groups of 3D points collected by laser scanners, stereo cameras, and drones. The data obtained from these devices has a wide range of applications. This article focuses on using point cloud data to verify planar structural elements of buildings. The article aims to present a methodology for analysing planar point cloud objects using 3D models in IFC format as reference data. The research includes processing data in IFC format, exporting geometric information, segmenting plane objects from a point cloud obtained by laser scanning, and analysing the data obtained. We provide the necessary graphic materials and code to enhance the content of the article. We also evaluate the results and create an outline for future work.

Advancing Our Understanding of Cloud Processes and Their Role in the Earth System through Cloud Object Tracking

Advancing Our Understanding of Cloud Processes and Their Role in the Earth System through Cloud Object Tracking

Convection Initiation Forecasting Using Synthetic Satellite Imagery from the Warn-on-Forecast System

Forecasting convection initiation (CI) has advanced greatly during the past decade through the use of high-resolution satellite observations and model output. One of the primary CI products used in forecast operations is based on GOES-16 visible and infrared imagery along with GLM lightning flash detections to determine the location of growing ice-containing cumulus clouds that are the precursor to developing thunderstorms. Another approach to CI forecasting that has recently become available is high frequency output from numerical weather prediction (NWP) models such as the Warn-on-Forecast System (WoFS). NWP model simulated composite reflectivity forecasts are one method used to determine when and where severe thunderstorms might develop. However, waiting for high reflectivity (> 40 dBZ) to be created within the NWP model limits the potential lead time available to forecasters when using WoFS output to anticipate areas where convection might form.Also, forecast reflectivity alone does not always give an indication of whether or not the precipitation developed by the NWP model is convective in nature. To address these limitations, this work applies a CI forecasting methodology developed for GOES satellite data on synthetic satellite imagery produced from WoFS output. Forecast cloud objects are tracked over a 10-min interval and CI forecasting parameters are applied to determine whether or not these cloud objects will continue to develop into organized thunderstorms.

DEVELOPMENT OF CLOUD DETECTION METHOD FOR CAS500-1 IMAGERY

Abstract. Clouds are typically characterized by high reflectance and low brightness temperature. They are generally classified as noise in optical land monitoring satellites. In particular, the presence of clouds has a decisive effect on the quality of follow-up data, so they must be detected. In this study, pre-processing was applied to effectively detect clouds while minimizing noise using CAS500-1 images with visible and near-infrared bands. First, the RGB color space is converted to the HSV color space. Next, a triangle thresholding method is applied to the value channel, which exhibits the highest correlation with pixel brightness, to extract bright objects. Then, the maximum likelihood method is applied to differentiate between bright objects and cloud candidate objects. Finally, threshold values for cloud detection are automatically determined to create initial cloud maps using the statistical values derived from the cloud candidate objects. We compared the results generated by the single thresholding method to verify the performance of the proposed method. As a result, the proposed method was able to detect clouds more accurately by considering the reflectance characteristics of each image. Moreover, except for cloud objects, the rest of the bright objects (white roofs, concrete roads, sand, etc.) were minimized. Our experiments showed high stability despite the absence of shortwave infrared and thermal infrared bands, which are effective for cloud detection.

3D point cloud object detection algorithm based on Transformer

In response to the difficulty in deploying anchor box based methods in 3D object detection due to the increase in spatial dimensions, this paper studies a point cloud object detection algorithm based on set prediction. This article proposes a Transformer based 3D point cloud object detection algorithm, and combines the characteristics of point clouds in autonomous driving scenarios to propose an improved spatial modulation attention and heat map initialization strategy for training acceleration and query initialization, achieving good detection performance in shallow networks. This article compares it with other algorithms on the KITTI dataset, and the results show that our algorithm has reached an advanced level in performance. We also conducted ablation experiments on the main components of the algorithm to verify the contribution of each module to the detection effect.

3D Object Detection Using Multiple-Frame Proposal Features Fusion.

Object detection is important in many applications, such as autonomous driving. While 2D images lack depth information and are sensitive to environmental conditions, 3D point clouds can provide accurate depth information and a more descriptive environment. However, sparsity is always a challenge in single-frame point cloud object detection. This paper introduces a two-stage proposal-based feature fusion method for object detection using multiple frames. The proposed method, called proposal features fusion (PFF), utilizes a cosine-similarity approach to associate proposals from multiple frames and employs an attention weighted fusion (AWF) module to merge features from these proposals. It allows for feature fusion specific to individual objects and offers lower computational complexity while achieving higher precision. The experimental results on the nuScenes dataset demonstrate the effectiveness of our approach, achieving an mAP of 46.7%, which is 1.3% higher than the state-of-the-art 3D object detection method.

Cross-domain point cloud completion for multi-class indoor incomplete objects via class-conditional GAN inversion

Indoor point clouds from real-world scans are often incomplete and sparse due to limited observation views and severe occlusion between objects. Point cloud completion can restore missing parts to ensure data integrity, so that subsequent tasks like point cloud classification, segmentation, object detection and three-dimensional (3D) reconstruction can obtain more accurate and complete geometric and semantic information. Existing point cloud completion methods focus more on synthetic data or require separate modeling on each class of objects, which struggle to meet the inherent requirements of the completion task for real-world indoor scenes containing multi-class objects with significant pose variations. In this research, we present an efficient class-conditional generative adversarial network (GAN) inversion framework for the task of recovering incomplete objects in real-world indoor scenes, in which we first train a class-conditional GAN to learn rich shape priors for multi-class objects from synthetic data domain and then find an optimal latent code in GAN’s latent space that gives a complete shape that best recovers the given incomplete input from real-world data domain. The performance of our proposed method is evaluated on both real-world and synthetic datasets. In comparison to ShapeInversion (Zhang et al., 2021), our method demonstrates an improvement in F-score by 3.7 and 2.0 percentage points while reducing the chamfer distance (CD) by 4.15 and 1.1 on ScanNet dataset and CRN dataset, respectively. The developed code will be publicly available at: https://github.com/CNU-DLandCV-lab/cgan-inversion.

Deep convective cloud system size and structure across the global tropics and subtropics

Abstract. A new database is constructed comprising millions of deep convective clouds that spans the global tropics and subtropics for the entire record of the MODIS instruments on the Terra and Aqua satellites. The database is a collection of individual cloud objects ranging from isolated convective cells to mesoscale convective cloud systems spanning hundreds of thousands of square kilometers in cloud area. By matching clouds in the database with the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) reanalysis dataset and microwave imager brightness temperatures from the Advanced Microwave Scanning Radiometer – EOS (AMSR-E) instrument, the database is designed to explore the relationships among the horizontal scale of cloud systems, the thermodynamic environment within which the cloud resides, the amount of aerosol in the environment, and indicators of the microphysical structure of the clouds. We find that the maximum values of convective available potential energy and vertical shear of the horizontal wind associated with a cloud impose a strong constraint on the size attained by a convective cloud system, although the relationship varies geographically. The cloud database provides a means of empirically studying the factors that determine the spatial structure and coverage of convective cloud systems, which are strongly related to the overall radiative forcing by cloud systems. The observed relationships between cloud system size and structure from this database can be compared with similar relationships derived from simulated clouds in atmospheric models to evaluate the representation of clouds and convection in weather forecast and climate projection simulations, including whether models exhibit the same relationships between the atmospheric environment and cloud system size and structure. Furthermore, the dataset is designed to probe the impacts of aerosols on the size and structure of deep convective cloud systems.

Oort cloud perturbations as a source of hyperbolic Earth impactors

The observation of interstellar objects 1I/’Oumuamua and 2I/Borisov suggests the existence of a larger population of smaller projectiles that impact our planet with unbound orbits. We analyze an asteroidal grazing meteor (FH1) recorded by the Finnish Fireball Network on October 23, 2022. FH1 displayed a likely hyperbolic orbit lying on the ecliptic plane with an estimated velocity excess of ∼0.7 kms−1 at impact. FH1 may either be an interstellar object, indicating a high-strength bias in this population, or an Oort cloud object, which would reinforce migration-based solar system models. Furthermore, under the calculated uncertainties, FH1 could potentially be associated with the passage of Scholz’s binary star system. Statistical evaluation of uncertainties in the CNEOS database and study of its hyperbolic fireballs reveals an anisotropic geocentric radiant distribution and low orbital inclinations, challenging the assumption of a randomly incoming interstellar population. Orbital integrations suggest that the event on March 9, 2017 (IM2) from CNEOS may have experienced gravitational perturbation during the Scholz fly-by, contingent upon velocity overestimation within the expected range. These findings suggest that apparent interstellar meteors may, in fact, be the result of accelerated meteoroid impacts caused by close encounters with massive objects within or passing through our solar system.

Observed Cloud Type‐Sorted Cloud Property and Radiative Flux Changes With the Degree of Convective Aggregation From CERES Data

AbstractCloud‐radiation interactions are a critical mechanism for convective self‐aggregation, especially the longwave radiative cooling of low clouds and environments. In this study, two data products from CERES observations combined with MERRA‐2 reanalysis are used to understand the changes of cloud properties and radiative fluxes by cloud type with the degree of convective aggregation at the 1000‐km scale, which is represented by the number of cloud objects (N), simple convective aggregation index (SCAI), modified SCAI (MCAI) or convective organization potential (COP). The changes with SCAI are similar to those with N as an index, agreeing with previous studies using grid‐averaged properties. For changes from weak to strong degrees of aggregation using N and SCAI, area fractions of middle‐ and high‐level cloud types decrease by up to 4% but those of low‐level cloud types increase by up to 2%, and more infrared radiation is emitted to space (2–8 W m−2) from optically thin cloud types but more solar radiation is reflected (2–4 W m−2) from optically‐thick cloud types. However, using COP (MCAI to lesser extent), area fractions of optically‐thick cloud types increase, which emit less infrared radiation and reflect more solar radiation, whereas the area fractions of low‐level clouds decrease. These results can be explained by greater expansion of cloud object sizes for COP than MCAI/SCAI as the degree of convective aggregation increases, which also explains the difference between SCAI and MCAI pertaining to the opposite changes of optically‐thick high‐level clouds. These results can have implications for understanding convective self‐aggregation.

ProxyMix: Proxy-based Mixup training with label refinery for source-free domain adaptation

Due to privacy concerns and data transmission issues, Source-free Unsupervised Domain Adaptation (SFDA) has gained popularity. It exploits pre-trained source models, rather than raw source data for target learning, to transfer knowledge from a labeled source domain to an unlabeled target domain. Existing methods solve this problem typically with additional parameters or noisy pseudo labels, and we propose an effective method named Proxy-based Mixup training with label refinery (ProxyMix) to avoid these drawbacks. To avoid additional parameters and leverages information in the source model, ProxyMix defines classifier weights as class prototypes and creates a class-balanced proxy source domain using nearest neighbors of the prototypes. To improve the reliability of pseudo labels, we further propose the frequency-weighted aggregation strategy to generate soft pseudo labels for unlabeled target data. Our strategy utilizes target features’ internal structure, increases weights of low-frequency class samples, and aligns the proxy and target domains using inter- and intra-domain mixup regularization. This mitigates the negative impact of noisy labels. Experiments on three 2D image and 3D point cloud object recognition benchmarks demonstrate that ProxyMix yields state-of-the-art performance for source-free UDA tasks.