LiDAR data generation has emerged as a promising solution to the high cost and limited scalability of real-world LiDAR sensing. Recent diffusion and rectified flow models have demonstrated strong capabilities in synthesizing realistic 3D point clouds; however, their iterative sampling procedures result in significant inference overhead. To address this, we focus on efficient few-step LiDAR generation for both unconditional and multi-modal conditional settings. Specifically, we propose an adaptive piecewise distillation strategy tailored for rectified flow-based LiDAR generation models, where the teacher model’s flow trajectory is adaptively segmented into consecutive intervals, and the student is trained only at the start of each interval to directly predict the velocity toward its endpoint. By sequentially sampling at the start timestep of each interval, our method enables fast few-step generation. Moreover, instead of uniform partitioning, we introduce an adaptive timestep selection strategy that chooses interval boundaries with minimal initial error, thereby reducing the complexity of distillation. Experimental results show that our method achieves comparable or superior performance to state-of-the-art methods in both unconditional and multi-modal conditional LiDAR generation, using only four sampling steps.

Understanding motion in dynamic environments is critical for autonomous driving, thereby motivating research on class-agnostic motion prediction. In this work, we investigate weakly and self-supervised class-agnostic motion prediction from LiDAR point clouds. Outdoor scenes typically consist of mobile foregrounds and static backgrounds, allowing motion understanding to be associated with scene parsing. Based on this observation, we propose a novel weakly supervised paradigm that replaces motion annotations with fully or partially annotated (1%, 0.1%) foreground/background masks for supervision. To this end, we develop a weakly supervised approach utilizing foreground/background cues to guide the self-supervised learning of motion prediction models. Since foreground motion generally occurs in non-ground regions, non-ground/ground masks can serve as an alternative to foreground/background masks, further reducing annotation effort. Leveraging non-ground/ground cues, we propose two additional approaches: a weakly supervised method requiring fewer (0.01%) foreground/background annotations, and a self-supervised method without annotations. Furthermore, we design a Robust Consistency-aware Chamfer Distance loss that incorporates multi-frame information and robust penalty functions to suppress outliers in self-supervised learning. Experiments show that our weakly and self-supervised models outperform existing self-supervised counterparts, and our weakly supervised models even rival some supervised ones. This demonstrates that our approaches effectively balance annotation effort and performance.

Radar-Camera depth estimation aims to predict dense and accurate metric depth by fusing input images and Radar data. Model efficiency is crucial for this task in pursuit of real-time processing on autonomous vehicles and robotic platforms. However, due to the sparsity of Radar returns, the prevailing methods adopt multi-stage frameworks with intermediate quasi-dense depth, which are time-consuming and not robust. To address these challenges, we propose TacoDepth, an efficient and accurate Radar-Camera depth estimation model with one-stage fusion. Specifically, the graph-based Radar structure extractor and the pyramid-based Radar fusion module are designed to capture and integrate the graph structures of Radar point clouds, delivering superior model efficiency and robustness without relying on the intermediate depth results. Moreover, TacoDepth can be flexible for different inference modes, providing a better balance of speed and accuracy. Extensive experiments are conducted to demonstrate the efficacy of our method. Compared with the previous state-of-the-art approach, TacoDepth improves depth accuracy and processing speed by 12.8% and 91.8%. Our work provides a new perspective on efficient Radar-Camera depth estimation.

Panoptic segmentation, which unifies semantic and instance segmentation into a single task, has witnessed considerable success on predefined tasks. However, traditional methods tend to struggle with catastrophic forgetting and poor generalization when learning from a continuous stream of new tasks. While continual learning aims to mitigate these challenges, our study reveals that existing continual panoptic segmentation (CPS) methods often suffer from efficiency or scalability issues. To address these limitations, we propose an efficient adaptation framework that incorporates attentive self-distillation and dual-decoder prediction fusion to efficiently preserve prior knowledge while facilitating model generalization. Specifically, we freeze the majority of model weights, enabling a shared forward pass between the teacher and student models during distillation. Attentive self-distillation then adaptively distills useful knowledge from the old classes without being distracted from non-object regions, which effectively enhances knowledge retention. Additionally, query-level fusion (QLF) is devised to seamlessly integrate the output of the dual decoders without incurring scale inconsistency. Our method achieves state-of-the-art performance on ADE20K and COCO benchmarks. Code is available at https://github.com/Ze-Yang/ADAPT.

High-resolution land surface temperature (LST) product holds significant importance in quantifying surface heat, monitoring climate change, assessing environmental health, and water resource management. Therefore, accurate LST retrieval improves our understanding of detailed thermal characteristics of the Earth’s surface. In this research, we developed an operational split-window algorithm for generationg high-resolution LST products from Gaofen5-02 (GF5-02) Visible and Infrared Multispectral Imager (VIMI) data. The coefficients of the split-window algorithm are simulated utilizing the MODTRAN 5.2 atmospheric radiative transfer model, with the global atmospheric profile library of SeeBor V5.0. The land surface emissivities in VIMI bands 11 and 12 are estimated using the ASTER global emissivity dataset (GED) based on the vegetation cover method. The GF5-02 VIMI LSTs are validated using in-situ data collected from the Huailai experiment site in China. Preliminary results show the accuracy of the GF5-02 LST products is satisfactory, exhibiting a bias of -0.03 K and a Root Mean Square Error (RMSE) of 2.42 K.

Thermal infrared (TIR) observation is the most widely used and accurate method for generating global land surface temperatures (LST) products. However, TIR signals are susceptible to cloud obscuration, and the affected region accounts for more than 50% of the global LST product. We referred to both analytical solution and optimization methods and proposed an all-weather LST reconstruction framework under the cloudy conditions for Chinese FengYun satellites (FY-3D MERSI-II and FY-4A AGRI) based on the principles of radiative transfer and surface energy balance. For FY-3D, the optimization method outperforms the analytical solution method, the overall bias (RMSE) of the estimated all-weather LST is 0.69 K (3.42 K) for the optimization method, while the overall bias (RMSE) of the analytical solution method is 1.33 K (3.52 K). For FY-4D, the validation results are similar for both methods, with a bias (RMSE) of 0.03 K (3.04 K) for the analytical solution method and a bias (RMSE) of -0.24 K (3.03 K) for the optimization method.

Sea Surface Temperature (SST) is a pivotal parameter in studying the energy balance and material exchange between the ocean and the atmosphere. Obtaining high-precision SST is of significant importance for a deep understanding of the dynamic changes in the ocean and atmospheric systems. This study utilized a nonlinear split-window algorithm for deriving 40m SST from Chinese Gaofen5-02 (GF5-02) Visible and Infrared Multispectral Imager (VIMI). The coefficients of the algorithm were simulated utilizing the MODTRAN 5.2 atmospheric radiative transfer model and the global atmospheric profile library of SeeBor V5.0. The retrieved VIMI SST was validated using iQuam SST measurements (in-situ measurements) and the MODIS SST products. The results of the cross-validation demonstrate a reasonable accuracy of the produced VIMI SST products, exhibiting a bias of 1.34 K and an RMSE of 2.68 K.

LST is a crucial climatic variable closely associated with surface energy and water balance. TIR observations are the most widely applied and accurate method for generating global LST products. Ensuring the accuracy of LST is of paramount importance. We calculated the LST from FY3D MERSI-II data using a generalized split-window algorithm and conducted validation. Initial validation was performed using data from sites in the OzFlux Observatory Network, HiWater Observatory Network, and SURFRAD Observatory Network from 2019 to 2020. The accuracy of the data at the sites of the three observation networks has a bias (RMSE) of 2.31 K (4.5 K) and -0.1 K (2.03 K) during daytime and nighttime, respectively. The surface temperature inversion results are slightly overestimated during the daytime and slightly underestimated during the nighttime, and the accuracy of the nighttime results is higher than that of the daytime results.

In this article, we investigate self-supervised 3D scene flow estimation and class-agnostic motion prediction on point clouds. A realistic scene can be well modeled as a collection of rigidly moving parts, therefore its scene flow can be represented as a combination of rigid motion of these individual parts. Building upon this observation, we propose to generate pseudo scene flow labels for self-supervised learning through piecewise rigid motion estimation, in which the source point cloud is decomposed into local regions and each region is treated as rigid. By rigidly aligning each region with its potential counterpart in the target point cloud, we obtain a region-specific rigid transformation to generate its pseudo flow labels. To mitigate the impact of potential outliers on label generation, when solving the rigid registration for each region, we alternately perform three steps: establishing point correspondences, measuring the confidence for the correspondences, and updating the rigid transformation based on the correspondences and their confidence. As a result, confident correspondences will dominate label generation, and a validity mask will be derived for the generated pseudo labels. By using the pseudo labels together with their validity mask for supervision, models can be trained in a self-supervised manner. Extensive experiments on FlyingThings3D and KITTI datasets demonstrate that our method achieves new state-of-the-art performance in self-supervised scene flow learning, without any ground truth scene flow for supervision, even performing better than some supervised counterparts. Additionally, our method is further extended to class-agnostic motion prediction and significantly outperforms previous state-of-the-art self-supervised methods on nuScenes dataset.

Predicting biomolecular interactions is significant for understanding biological systems. Most existing methods for link prediction are based on graph convolution. Although graph convolution methods are advantageous in extracting structure information of biomolecular interactions, two key challenges still remain. One is how to consider both the immediate and high-order neighbors. Another is how to reduce noise when aggregating high-order neighbors. To address these challenges, we propose a novel method, called mixed high-order graph convolution with filter network via LSTM and channel attention (HGLA), to predict biomolecular interactions. Firstly, the basic and high-order features are extracted respectively through the traditional graph convolutional network (GCN) and the two-layer Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixing (MixHop). Secondly, these features are mixed and input into the filter network composed of LayerNorm, SENet and LSTM to generate filtered features, which are concatenated and used for link prediction. The advantages of HGLA are: 1) HGLA processes high-order features separately, rather than simply concatenating them; 2) HGLA better balances the basic features and high-order features; 3) HGLA effectively filters the noise from high-order neighbors. It outperforms state-of-the-art networks on four benchmark datasets.

Understanding the motion behavior of dynamic environments is vital for autonomous driving, leading to increasing attention in class-agnostic motion prediction in LiDAR point clouds. Outdoor scenes can often be decomposed into mobile foregrounds and static backgrounds, which enables us to associate motion understanding with scene parsing. Based on this observation, we study a novel weakly supervised motion prediction paradigm, where fully or partially (1 %, 0.1%) annotated foreground/background binary masks are used for supervision, rather than using expensive motion annotations. To this end, we propose a two-stage weakly supervised approach, where the segmentation model trained with the incomplete binary masks in Stage1 will facilitate the self-supervised learning of the motion prediction network in Stage2 by estimating possible moving foregrounds in advance. Furthermore, for robust self-supervised motion learning, we design a Consistency-aware Chamfer Distance loss by exploiting multi-frame information and explicitly suppressing potential outliers. Comprehensive experiments show that, with fully or partially binary masks as supervision, our weakly supervised models surpass the self-supervised models by a large margin and perform on par with some supervised ones. This further demonstrates that our approach achieves a good compromise between annotation effort and performance.

Continual semantic segmentation (CSS) aims to extend an existing model to tackle unseen tasks while retaining its old knowledge. Naively fine-tuning the old model on new data leads to catastrophic forgetting. A common solution is knowledge distillation (KD), where the output distribution of the new model is regularized to be similar to that of the old model. However, in CSS, this is challenging because of the background shift issue. Existing KD-based CSS methods continue to suffer from confusion between the background and novel classes since they fail to establish a reliable class correspondence for distillation. To address this issue, we propose a new label-guided knowledge distillation (LGKD) loss, where the old model output is expanded and transplanted (with the guidance of the ground truth label) to form a semantically appropriate class correspondence with the new model output. Consequently, the useful knowledge from the old model can be effectively distilled into the new model without causing confusion. We conduct extensive experiments on two prevailing CSS benchmarks, Pascal-VOC and ADE20K, where our LGKD significantly boosts the performance of three competing methods, especially on novel mIoU by up to +76%, setting new state-of-the-art. Finally, to further demonstrate its generalization ability, we introduce the first CSS benchmark for 3D point cloud based on ScanNet, along with several re-implemented baselines for comparison. Experiments show that LGKD is versatile in both 2D and 3D modalities without requiring ad hoc design. Codes are available at https://github.com/Ze-Yang/LGKD.

Instance segmentation on 3D point clouds has been attracting increasing attention due to its wide applications, especially in scene understanding areas. However, most existing methods operate on fully annotated data while manually preparing ground-truth labels at point-level is very cumbersome and labor-intensive. To address this issue, we propose a novel weakly supervised method RWSeg that only requires labeling one object with one point. With these sparse weak labels, we introduce a unified framework with two branches to propagate semantic and instance information respectively to unknown regions using self-attention and a cross-graph random walk method. Specifically, we propose a Cross-graph Competing Random Walks (CRW) algorithm that encourages competition among different instance graphs to resolve ambiguities in closely placed objects, improving instance assignment accuracy. RWSeg generates high-quality instance-level pseudo labels. Experimental results on ScanNet-v2 and S3DIS datasets show that our approach achieves comparable performance with fully-supervised methods and outperforms previous weakly-supervised methods by a substantial margin.

The Raft consensus protocol naturally fits time series databases, owing to the resemblance between its continuous log and the time series data. While the serialization of appending entries reduces the state space for ease design and implementation, it blocks the subsequent requests and thus limits the parallelism and throughput of Raft. Intuitively, once an entry arrives the follower, we may notice the leader and the client to unblock the subsequent as early, rather than waiting for its appending and committing. In this way, more requests can be processed in parallel, and thus the throughput increases, essential for IoT applications often with vast sensors and fast data ingestion. Of course, with higher parallelism, the risk of persistence for in-processing entries increases. It is a worthwhile trade-off in the IoT scenario since tiny data loss during leader failure is more acceptable than shutting out most data due to a low throughput. Our Non-Blocking Raft (NB-Raft) is implemented as the consensus protocol of Apache IoTDB, a commodity time series database management system, supporting various applications in Alibaba Cloud. Extensive evaluation shows that the throughput is improved by about 30% using our NB-Raft compared to the original Raft, a considerable amount of further data saved.

Land surface temperature (LST) is a vital parameter for studying global ecological, climatic, and environmental changes. A variety of regional and global scale LST products have been produced based on satellite remote sensing. Therefore, reliable retrieval accuracy is crucial for the application of LST products. A ground measurement processing method for medium and high spatial resolution LST products is proposed in this paper, to solve the spatial scale issues between the ground radiometer's field-of-view and the satellite pixel in T-based validation. The method is divided into two steps: LST quality control and spatial scale transformation. The Thermal Airborne Spectrographic Imager (TASI) data was used to evaluate the accuracy of the method. The results showed that this method can improve the reliability of the ground LST measurements.

The goal of 3D pose transfer is to transfer the pose from the source mesh to the target mesh while preserving the identity information (e.g., face, body shape) of the target mesh. Deep learning-based methods improved the efficiency and performance of 3D pose transfer. However, most of them are trained under the supervision of the ground truth, whose availability is limited in real-world scenarios. In this work, we present X-DualNet, a simple yet effective approach that enables unsupervised 3D pose transfer. In X-DualNet, we introduce a generator G which contains correspondence learning and pose transfer modules to achieve 3D pose transfer. We learn the shape correspondence by solving an optimal transport problem without any key point annotations and generate high-quality meshes with our elastic instance normalization (ElaIN) in the pose transfer module. With G as the basic component, we propose a cross consistency learning scheme and a dual reconstruction objective to learn the pose transfer without supervision. Besides that, we also adopt an as-rigid-as-possible deformer in the training process to fine-tune the body shape of the generated results. Extensive experiments on human and animal data demonstrate that our framework can successfully achieve comparable performance as the state-of-the-art supervised approaches.

In real-world LiDAR-based applications, data is generated in the form of 3D point cloud sequences or 4D point clouds. However, the topic of semantic segmentation on 4D point clouds is under-investigated and existing methods are still not able to achieve satisfactory performance to meet the requirement for real-world applications. The temporal information across different point clouds plays an important role in dynamic scene understanding, which is not well explored in existing work. In this paper, we focus on exploring effective temporal information across two consecutive point clouds for semantic segmentation on point cloud sequences. To this end, we design three novel modules to enhance the features of target frames by extracting different temporal information in the local regions and global regions. Experimental results on SemanticKITTI and SemanticPOSS demonstrate that our method achieves superior performance in 4D semantic segmentation by utilizing temporal information.

Land surface temperature (LST) is a vital parameter for studying global ecological, climatic, and environmental changes. Although various LST retrieval algorithms have been proposed, including split-window (SW), dual-window (DW), three-channel (TC), and dual-angle (DA) algorithms, few studies have compared these algorithms using the same satellite observations. The Sea and Land Surface Temperature Radiometer (SLSTR) onboard Sentinel-3A provides a unique opportunity to conduct this comparison owing to its dual-angle viewing capability and multiple thermal infrared (TIR) and mid-infrared (MIR) channels. Here, we implemented two SW algorithms, one DW algorithm, two TC algorithms and one DA algorithm for the SLSTR data. The LST retrievals from these six algorithms were validated, along with the SLSTR operational LST product based on an emissivity-implicit SW algorithm. Temperature-based and radiance-based validation methods were used to evaluate different LST retrievals across different land cover types. The results indicated that the proposed SW algorithm had the highest accuracy, followed by the Pérez-Planells SW and the official algorithms. The overall root-mean-square errors (RMSEs) of these three SW algorithms were 1.42 K, 1.79 K and 2.05 K, respectively. The three algorithms involving the MIR channel (one DW and two TC algorithms) were more suitable for nighttime LST retrieval and had similar performances to the three SW algorithms, with a nighttime RMSE of approximately 1.36 K. The LST retrieval accuracy of the DA algorithm had the highest uncertainty and was closely related to the angular variation in surface emissivity and brightness temperature. The findings of this study contribute to a better understanding of the different LST retrieval algorithms and facilitate potential improvements in the official LST retrieval algorithm for SLSTR.

Land surface temperature (LST) is an important parameter that characterizes the energy balance of the land surface, and it is widely used in various research fields. This paper proposes an operational split-window (SW) algorithm for use with the Chinese Fengyun-3 (FY-3) series satellite data, with the purpose of generating long-term global LST products. The algorithm primarily involves three steps. First, the brightness temperatures of the FY-3 Visible and Infra-Red Radiometer (VIRR) were recalibrated using historical recalibration coefficients to improve the accuracy of the absolute radiometric calibration. Second, daily dynamic emissivity maps were estimated using the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) global emissivity dataset (GED) and vegetation/snow cover products based on the vegetation cover method. Finally, the coefficients of the SW algorithm were simulated using MODTRAN 5 combined with the SeeBor V5.0 atmospheric profile library and ASTER spectral library, and then the coefficients were stratified by the view zenith angle and atmospheric water vapor content to improve the fitting accuracy. The proposed SW algorithm was integrated into the MUlti-source data SYnergized Quantitative (MUSYQ) remote sensing production system to then generate FY-3 VIRR LST products. Ten land surface sites from the HiWATER and SURFRAD networks and nine water surface sites from the National Data Buoy Center (NDBC) were used to evaluate the accuracy of the FY-3 VIRR LST products. The results demonstrated that the accuracy of the historical recalibration coefficients of the FY-3A/B VIRR is higher than that of the operational calibration coefficients for LST retrieval. The evaluation results revealed that the FY-3A VIRR LST products (2009-2013) had a bias of 0.13 K and an RMSE of 2.77 K, and the FY-3B VIRR LST products (2011-2020) had a bias of -0.07 K and an RMSE of 2.83 K. These results demonstrate that the proposed operational SW algorithm has reasonable accuracy and can be used to produce global LST products from the FY-3 VIRR data.

Few-shot object detection (FSOD), which aims at learning a generic detector that can adapt to unseen tasks with scarce training samples, has witnessed consistent improvement recently. However, most existing methods ignore the efficiency issues, e.g., high computational complexity and slow adaptation speed. Notably, efficiency has become an increasingly important evaluation metric for few-shot techniques due to an emerging trend toward embedded AI. To this end, we present an efficient pretrain-transfer framework (PTF) baseline with no computational increment, which achieves comparable results with previous state-of-the-art (SOTA) methods. Upon this baseline, we devise an initializer named knowledge inheritance (KI) to reliably initialize the novel weights for the box classifier, which effectively facilitates the knowledge transfer process and boosts the adaptation speed. Within the KI initializer, we propose an adaptive length re-scaling (ALR) strategy to alleviate the vector length inconsistency between the predicted novel weights and the pretrained base weights. Finally, our approach not only achieves the SOTA results across three public benchmarks, i.e., PASCAL VOC, COCO and LVIS, but also exhibits high efficiency with $1.8-100\times $ faster adaptation speed against the other methods on COCO/LVIS benchmark during few-shot transfer. To our best knowledge, this is the first work to consider the efficiency problem in FSOD. We hope to motivate a trend toward powerful yet efficient few-shot technique development. The codes are publicly available at https://github.com/Ze-Yang/Efficient-FSOD.

Existing point cloud segmentation methods require a large amount of annotated data, especially for the outdoor point cloud scene. Due to the complexity of the outdoor 3D scenes, manual annotations on the outdoor point cloud scene are time-consuming and expensive. In this paper, we study how to achieve scene understanding with limited annotated data. Treating 100 consecutive frames as a sequence, we divide the whole dataset into a series of sequences and annotate only 0.1% points in the first frame of each sequence to reduce the annotation requirements. This leads to a total annotation budget of 0.001%. We propose a novel temporal-spatial framework for effective weakly supervised learning to generate high-quality pseudo labels from these limited annotated data. Specifically, the frame-work contains two modules: an matching module in temporal dimension to propagate pseudo labels across different frames, and a graph propagation module in spatial dimension to propagate the information of pseudo labels to the entire point clouds in each frame. With only 0.001% annotations for training, experimental results on both SemanticKITTI and SemanticPOSS shows our weakly supervised two-stage framework is comparable to some existing fully supervised methods. We also evaluate our framework with 0.005% initial annotations on SemanticKITTI, and achieve a result close to fully supervised backbone model.

In this work, we focus on scene flow learning on point clouds in a self-supervised manner. A real-world scene can be well modeled as a collection of rigidly moving parts, therefore its scene flow can be represented as a combination of rigid motion of each part. Inspired by this observation, we propose to generate pseudo scene flow for self-supervised learning based on piecewise rigid motion estimation, in which the source point cloud is decomposed into a set of local regions and each region is treated as rigid. By rigidly aligning each region with its potential counterpart in the target point cloud, we obtain a region-specific rigid transformation to represent the flow, which together constitutes the pseudo scene flow labels of the entire scene to enable network training. Compared with most existing approaches relying on point-wise similarities for scene flow approximation, our method explicitly enforces region-wise rigid alignments, yielding locally rigid pseudo scene flow labels. We demonstrate the effectiveness of our self-supervised learning method on FlyingThings3D and KITTI datasets. Comprehensive experiments show that our method achieves new state-of-the-art performance in self-supervised scene flow learning, without any ground truth scene flow for supervision, even outperforming some super-vised counterparts.

High-resolution land surface temperature (LST) retrieval is a hot research topic in recent ten years, and the development of various high-resolution satellite sensors provides a data basis for this study. The Visual and Infrared Multispectral Imager (VIMI) on Gaofen5-02 (GF5-02) satellite provides 40m spatial resolution thermal infrared data ranging from 8µm to 12.5µm, including four thermal infrared channels. In this paper, we developed a split-window algorithm for retrieving LST from VIMI data. First, the two thermal infrared channels 11 and 12 of VIMI are cross-calibrated using MODIS bands 31 and 32, and then high-resolution LST was derived using the generalized split-window algorithm. The GF5-02 LST was cross-validated with the MODIS MOD21 LST products, the preliminary results indicate that GF5-02 LST shows a reasonable accuracy, with a mean bias of 0.05 K and a mean RMSE of 3.29 K.

Scene flow in 3D point clouds plays an important role in understanding dynamic environments. Although significant advances have been made by deep neural networks, the performance is far from satisfactory as only per-point translational motion is considered, neglecting the constraints of the rigid motion in local regions. To address the issue, we propose to introduce the motion consistency to force the smoothness among neighboring points. In addition, constraints on the rigidity of the local transformation are also added by sharing unique rigid motion parameters for all points within each local region. To this end, a high-order CRFs based relation module (Con-HCRFs) is deployed to explore both point-wise smoothness and region-wise rigidity. To empower the CRFs to have a discriminative unary term, we also introduce a position-aware flow estimation module to be incorporated into the Con-HCRFs. Comprehensive experiments on FlyingThings3D and KITTI show that our proposed framework (HCRF-Flow) achieves state-of-the-art performance and significantly outperforms previous approaches substantially.

Due to the scarcity of annotated scene flow data, self-supervised scene flow learning in point clouds has attracted increasing attention. In the self-supervised manner, establishing correspondences between two point clouds to ap-proximate scene flow is an effective approach. Previous methods often obtain correspondences by applying point-wise matching that only takes the distance on 3D point co-ordinates into account, introducing two critical issues: (1) it overlooks other discriminative measures, such as color and surface normal, which often bring fruitful clues for ac-curate matching; and (2) it often generates sub-par performance, as the matching is operated in an unconstrained situation, where multiple points can be ended up with the same corresponding point. To address the issues, we formulate this matching task as an optimal transport problem. The output optimal assignment matrix can be utilized to guide the generation of pseudo ground truth. In this optimal transport, we design the transport cost by considering multiple descriptors and encourage one-to-one matching by mass equality constraints. Also, constructing a graph on the points, a random walk module is introduced to encourage the local consistency of the pseudo labels. Comprehensive experiments on FlyingThings3D and KITTI show that our method achieves state-of-the-art performance among self-supervised learning methods. Our self-supervised method even performs on par with some supervised learning approaches, although we do not need any ground truth flow for training.

Few-shot learning aims to adapt knowledge learned from previous tasks to novel tasks with only a limited amount of labeled data. Research literature on few-shot learning exhibits great diversity, while different algorithms often excel at different few-shot learning scenarios. It is therefore tricky to decide which learning strategies to use under different task conditions. Inspired by the recent success in Automated Machine Learning literature (AutoML), in this paper, we present Meta Navigator, a framework that attempts to solve the aforementioned limitation in few-shot learning by seeking a higher-level strategy and proffer to automate the selection from various few-shot learning designs. The goal of our work is to search for good parameter adaptation policies that are applied to different stages in the network for few-shot classification. We present a search space that covers many popular few-shot learning algorithms in the literature, and develop a differentiable searching and decoding algorithm based on meta-learning that supports gradient-based optimization. We demonstrate the effectiveness of our searching-based method on multiple benchmark datasets. Extensive experiments show that our approach significantly outperforms baselines and demonstrates performance advantages over many state-of-the-art methods.

In this study, a split-window(SW) algorithm is tested to retrieve the land surface temperature (LST) from nighttime mid-infrared MODIS data. At first, the SW algorithm's coefficients were derived using MODTRAN simulations with the TIGR atmospheric profile database. Then, the input emissivities of the SW algorithm were directly calculated using the MODIS MYD11B1 products. At last, the LST was retrieved using the Aqua MODIS bands 22 and 23 data. The retrieved LSTs were compared with the MODIS MYD11A1 products and validated using ground measurements collected from four ground sites in northwest China. The validation results indicate that the developed SW algorithm provides better accuracy than the MYD11A1 LST products, with a mean bias of −0.76K and a mean root-mean-square error (RMSE) of 1.3K. Experiments show that the split-window algorithm is also applicable in the mid-infrared channel at night, and can produce accurate land surface temperature results.

In this study, we have developed an operational split-window algorithm for retrieving the land surface temperature (LST) from Chinese FengYun-3 (FY-3) series satellite data, with the purpose of generating long-term FY-3 LST products from 2009 to 2020. The refined generalized split-window (GSW) algorithm was selected and the coefficients of the algorithm were simulated using radiative transfer model MODTRAN 5.2 and Seebore v5.0 atmospheric profile database. The land surface emissivities (LSE) in the two SW channels were calculated using the ASTER Global Emissivity Database (GED), vegetation cover product and snow cover product based on the vegetation cover method. The developed FY-3 GSW algorithm was implemented in a MUlti-source data SYnergized Quantitative (MuSyQ) remote sensing product production system. The FY-3A and FY-3B LST products in northwest China were produced for 2013 and 2014, respectively, and the results were evaluated using ground measurements collected in four barren surface sites in the Heihe river basin. Both level 1 and recalibrated VIRR data were used for retrieving the LSTs. The results showed that the historical recalibration coefficients of the VIRR data can improve the accuracy of the LST retrievals.

3D pose transfer is one of the most challenging 3D generation tasks. It aims to transfer the pose of a source mesh to a target mesh and keep the identity (e.g., body shape) of the target mesh. Some previous works require key point annotations to build reliable correspondence between the source and target meshes, while other methods do not consider any shape correspondence between sources and targets, which leads to limited generation quality. In this work, we propose a correspondence-refinement network to achieve the 3D pose transfer for both human and animal meshes. The correspondence between source and target meshes is first established by solving an optimal transport problem. Then, we warp the source mesh according to the dense correspondence and obtain a coarse warped mesh. The warped mesh will be better refined with our proposed Elastic Instance Normalization, which is a conditional normalization layer and can help to generate high-quality meshes. Extensive experimental results show that the proposed architecture can effectively transfer the poses from source to target meshes and produce better results with satisfied visual performance than state-of-the-art methods.

In this study, two collection 6 (C6) Moderate Resolution Imaging Spectroradiometer (MODIS) level-2 land surface temperature (LST) products (MYD11_L2 and MYD21_L2) from the Aqua satellite were evaluated using temperature-based (T-based) and radiance-based (R-based) validation methods over barren surfaces in Northwestern China. The ground measurements collected at four barren surface sites from June 2012 to September 2018 during the Heihe Watershed Allied Telemetry Experimental Research (HiWATER) experiment were used to perform the T-based evaluation. Ten sand dune sites were selected in six large deserts in Northwestern China to carry out an R-based validation from 2012 to 2018. The T-based validation results indicate that the C6 MYD21 LST product has a better accuracy than the C6 MYD11 product during both daytime and nighttime. The LST is underestimated by the C6 MYD11 products at the four T-based sites during the daytime, with a mean bias of -2.82 K and a mean RMSE of 3.82 K, whereas the MYD21 LST product has a mean bias and RMSE of -0.51 and 2.53 K, respectively. The LST is also underestimated at night by the C6 MYD11 products at the four T-based sites, with a mean bias of -1.40 K and a mean RMSE of 1.72 K, whereas the MYD21 LST product has a mean bias and RMSE of 0.23 and 1.01 K, respectively. For the R-based validation, the MYD11 results are associated with large negative biases during both daytime and nighttime at three sand dune sites and biases within 1 K at the other seven sites, whereas the MYD21 results are more consistent at all ten sand dune sites, with a mean bias of 0.45 and 0.70 K for daytime and nighttime, respectively. The emissivities for these two products in MODIS bands 31 and 32 were compared with each other and then compared with the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) emissivity and laboratory emissivity. The results indicate that the emissivities in MODIS bands 31 and 32 of MYD11 at the four T-based and three of the R-based validation sites are overestimated and result in LST underestimation, whereas the emissivities of MYD21 are more consistent with the laboratory emissivity. Besides, an experiment was carried out to demonstrate that the physically retrieved dynamic emissivity of the MYD21 product can be utilized to improve the accuracy of the split-window (SW) algorithm for barren surfaces, making it a valuable data source for retrieving LST from different remote sensing data.

JSON is a very popular data format in many applications in Web and enterprise. Recently, many data analytical systems support the loading and querying JSON data. However, JSON parsing can be costly, which dominates the execution time of querying JSON data. Many previous studies focus on building efficient parsers to reduce this parsing cost, and little work has been done on how to reduce the occurrences of parsing. In this paper, we start with a study with a real production workload in Alibaba, which consists of over 3 million queries on JSON. Our study reveals significant temporal and spatial correlations among those queries, which result in massive redundant parsing operations among queries. Instead of repetitively parsing the JSON data, we propose to develop a cache system named Maxson for caching the JSON query results (the values evaluated from JSONPath) for reuse. Specifically, we develop effective machine learning-based predictor with combining LSTM (long shortterm memory) and CRF (conditional random field) to determine the JSONPaths to cache given the space budget. We have implemented Maxson on top of SparkSQL. We experimentally evaluate Maxson and show that 1) Maxson is able to eliminate the most of duplicate JSON parsing overhead, 2) Maxson improves end-to-end workload performance by 1.5-6.5×.

In this paper, in order to produce long term fully global land surface temperature (LST) product, the generalized split-window (GSW) algorithm and dual-window (DW) algorithm was used to retrieve LST from different geostationary (GEO) satellite data, including the FY-2E/4A, MTSAT-2/Himawari-8, MSG2, and GOES13/15. First, the coefficients of the GSW and DW algorithm were obtained from a simulation database constructed using the MODTRAN 5.2 and the SeeBor V5.0 atmospheric profile database. Second, the emissivity was estimated using the vegetation cover method, with the bare soil component emissivity derived from the ASTER global emissivity dataset (GED). Finally, the LST results of FY-4A AGRI and Himawari-8 AHI were retrieved and cross-validated. The results show that the LST algorithms developed in this work are capable of generating accurate high temporal resolution LST retrieval from global GEO satellite data.

In this study, we present an operational algorithm to retrieve the land surface temperature (LST) from MODIS thermal infrared data using physically retrieved emissivity product. This algorithm was implemented in the MUlti-source data SYnergized Quantitative (MuSyQ) remote sensing product system. First, the emissivity in the MODIS two split-window channels was estimated using the vegetation cover method, with the bare soil component emissivity derived from the ASTER global emissivity dataset. Then, the LST was retrieved using a modified generalized split-window algorithm. The MuSyQ MODIS LST product and the C6 MxD11 LST product were evaluated using ground measurements collected from four barren surface sites in Northwest China during the Heihe Watershed Allied Telemetry Experimental Research (HiWATER) experiment. The evaluation results indicate that the MuSyQ LST products provide better accuracy than the C6 MxD11 product during both daytime and nighttime at all four sites. This study demonstrates that physically retrieved emissivity products are a useful source for LST retrieval over barren surfaces.

In this study, to improve the accuracy of land surface temperature (LST) products over barren surfaces, we present an operational algorithm to retrieve the LST from Moderate-Resolution Imaging Spectroradiometer (MODIS) thermal infrared data using physically retrieved emissivity products. The LST algorithm involved two steps. First, the emissivity in the two MODIS split-window (SW) channels was estimated using the vegetation cover method, with the bare soil component emissivity derived from the ASTER global emissivity data set. Then, the LST was retrieved using a modified generalized SW algorithm. This algorithm was implemented in the MUlti-source data SYnergized Quantitative (MuSyQ) remote sensing product system. The MuSyQ MODIS LST product and the Collection 6 MODIS LST product (MxD11_L2) were compared and validated using ground measurements collected from four barren surface sites in Northwest China during the Heihe Watershed Allied Telemetry Experimental Research (HiWATER) experiment from June 2012 to December 2015. In total, 2268 and 2715 clear-sky samples were used in the validation for Terra and Aqua, respectively. The evaluation results indicate that the MuSyQ LST products provide better accuracy than the C6 MxD11 product during both daytime and nighttime at all four sites. For the daytime results, the LST is underestimated by the C6 MxD11 products at all four sites, with a mean bias of -1.78 and -2.86 K and a mean root-mean-square error (RMSE) of 3.16 and 3.94 K for Terra and Aqua, respectively, whereas the mean biases of the MuSyQ LST products are within 1 K, with a mean bias of -0.26 and -1.03 K and a mean RMSE of 2.45 and 2.71 K for Terra and Aqua, respectively. For the nighttime results, the LST is also underestimated by the C6 MxD11 products at all four sites, with a mean bias of -1.60 and -1.26 K and a mean RMSE of 1.93 and 1.60 K for Terra and Aqua, respectively, whereas the mean biases of the MuSyQ LST products are 0.16 and 0.58 K and the mean RMSEs are 1.12 and 1.25 K for Terra and Aqua, respectively. The results indicate that the underestimation of the C6 MxD11 LST product at all four sites mainly results from the overestimation of the emissivities in MODIS bands 31 and 32. This study demonstrates that physically retrieved emissivity products are a useful source for LST retrieval over barren surfaces and can be used to improve the accuracy of global LST products.

In this paper we study the problem of monocular relative depth perception in the wild. We introduce a simple yet effective method to automatically generate dense relative depth annotations from web stereo images, and propose a new dataset that consists of diverse images as well as corresponding dense relative depth maps. Further, an improved ranking loss is introduced to deal with imbalanced ordinal relations, enforcing the network to focus on a set of hard pairs. Experimental results demonstrate that our proposed approach not only achieves state-of-the-art accuracy of relative depth perception in the wild, but also benefits other dense per-pixel prediction tasks, e.g., metric depth estimation and semantic segmentation.

To improve the spatial resolution of the global aerosol optical depth (AOD) distribution, a new method for AOD retrieval is proposed over land in this paper. A monthly Global Land Surface Reflectance Database (GLSRD) was constructed using the long time serious of MODIS surface reflectance product (MOD09A1) and used for the surface reflectance estimation over land for AOD retrieval. A seasonal Global Land Aerosol Type Database (GLATD) was also built based on the MODIS aerosol product (MOD04) and used for providing the aerosol types over land in AOD retrieval. Thus, the AOD global dataset with 1 km resolution was produced and was validated against with the AERONET ground-based AOD measurements located in the global 172 stations over land. Results showed that the AOD retrievals are highly consistent with AERONET AODs and showed an overall better precision over both dark and bright areas.

Cloud detection is a key processing step before extracting information of earth surface from the earth observation data. Lots of schemes have been developed for cloud detection, static threshold method is the main method that is widely used in cloud detection. However, for the huge difference between different land objects, it is much difficult to find a proper threshold to detect the cloudy pixel from clear sky, especially, when the land covered by the thin or broken cloud. Therefore, a dynamic threshold cloud detection algorithm was proposed in this paper to improve the cloud detection. A priori monthly surface reflectance database was constructed using MODIS surface reflectance products and used to estimate the surface reflectance for dynamic threshold determination. Dynamic thresholds were determined by the simulation relationships between the apparent reflectance and the surface reflectance under clear conditions with 6S model. MODIS and Landsat 8 OLI data were selected to perform the experiments. Results showed that this new algorithm demonstrated better detection results of different cloud types over different land types.