Journal of Computer Science and Technology最新文献

Traditional neural radiance fields for rendering novel views require intensive input images and pre-scene optimization, which limits their practical applications. We propose a generalization method to infer scenes from input images and perform high-quality rendering without pre-scene optimization named SG-NeRF (Sparse-Input Generalized Neural Radiance Fields). Firstly, we construct an improved multi-view stereo structure based on the convolutional attention and multi-level fusion mechanism to obtain the geometric features and appearance features of the scene from the sparse input images, and then these features are aggregated by multi-head attention as the input of the neural radiance fields. This strategy of utilizing neural radiance fields to decode scene features instead of mapping positions and orientations enables our method to perform cross-scene training as well as inference, thus enabling neural radiance fields to generalize for novel view synthesis on unseen scenes. We tested the generalization ability on DTU dataset, and our PSNR (peak signal-to-noise ratio) improved by 3.14 compared with the baseline method under the same input conditions. In addition, if the scene has dense input views available, the average PSNR can be improved by 1.04 through further refinement training in a short time, and a higher quality rendering effect can be obtained.

The advent of Big Data has led to the rapid growth in the usage of parallel clustering algorithms that work over distributed computing frameworks such as MPI, MapReduce, and Spark. An important step for any parallel clustering algorithm is the distribution of data amongst the cluster nodes. This step governs the methodology and performance of the entire algorithm. Researchers typically use random, or a spatial/geometric distribution strategy like kd-tree based partitioning and grid-based partitioning, as per the requirements of the algorithm. However, these strategies are generic and are not tailor-made for any specific parallel clustering algorithm. In this paper, we give a very comprehensive literature survey of MPI-based parallel clustering algorithms with special reference to the specific data distribution strategies they employ. We also propose three new data distribution strategies namely Parameterized Dimensional Split for parallel density-based clustering algorithms like DBSCAN and OPTICS, Cell-Based Dimensional Split for dGridSLINK, which is a grid-based hierarchical clustering algorithm that exhibits efficiency for disjoint spatial distribution, and Projection-Based Split, which is a generic distribution strategy. All of these preserve spatial locality, achieve disjoint partitioning, and ensure good data load balancing. The experimental analysis shows the benefits of using the proposed data distribution strategies for algorithms they are designed for, based on which we give appropriate recommendations for their usage.

Automated test generation tools enable test automation and further alleviate the low efficiency caused by writing hand-crafted test cases. However, existing automated tools are not mature enough to be widely used by software testing groups. This paper conducts an empirical study on the state-of-the-art automated tools for Java, i.e., EvoSuite, Randoop, JDoop, JTeXpert, T3, and Tardis. We design a test workflow to facilitate the process, which can automatically run tools for test generation, collect data, and evaluate various metrics. Furthermore, we conduct empirical analysis on these six tools and their related techniques from different aspects, i.e., code coverage, mutation score, test suite size, readability, and real fault detection ability. We discuss about the benefits and drawbacks of hybrid techniques based on experimental results. Besides, we introduce our experience in setting up and executing these tools, and summarize their usability and user-friendliness. Finally, we give some insights into automated tools in terms of test suite readability improvement, meaningful assertion generation, test suite reduction for random testing tools, and symbolic execution integration.

Offline handwritten formula recognition is a challenging task due to the variety of handwritten symbols and two-dimensional formula structures. Recently, the deep neural network recognizers based on the encoder-decoder framework have achieved great improvements on this task. However, the unsatisfactory recognition performance for formulas with long LATEX strings is one shortcoming of the existing work. Moreover, lacking sufficient training data also limits the capability of these recognizers. In this paper, we design a multimodal dependence attention (MDA) module to help the model learn visual and semantic dependencies among symbols in the same formula to improve the recognition performance of the formulas with long LATEX strings. To alleviate overfitting and further improve the recognition performance, we also propose a new dataset, Handwritten Formula Image Dataset (HFID), which contains 25 620 handwritten formula images collected from real life. We conduct extensive experiments to demonstrate the effectiveness of our proposed MDA module and HFID dataset and achieve state-of-the-art performances, 63.79% and 65.24% expression accuracy on CROHME 2014 and CROHME 2016, respectively.

Novel artificial intelligence (AI) technology has expedited various scientific research, e.g., cosmology, physics, and bioinformatics, inevitably becoming a significant category of workload on high-performance computing (HPC) systems. Existing AI benchmarks tend to customize well-recognized AI applications, so as to evaluate the AI performance of HPC systems under the predefined problem size, in terms of datasets and AI models. However, driven by novel AI technology, most of AI applications are evolving fast on models and datasets to achieve higher accuracy and be applicable to more scenarios. Due to the lack of scalability on the problem size, static AI benchmarks might be under competent to help understand the performance trend of evolving AI applications on HPC systems, in particular, the scientific AI applications on large-scale systems. In this paper, we propose a scalable evaluation methodology (SAIH) for analyzing the AI performance trend of HPC systems with scaling the problem sizes of customized AI applications. To enable scalability, SAIH builds a set of novel mechanisms for augmenting problem sizes. As the data and model constantly scale, we can investigate the trend and range of AI performance on HPC systems, and further diagnose system bottlenecks. To verify our methodology, we augment a cosmological AI application to evaluate a real HPC system equipped with GPUs as a case study of SAIH. With data and model augment, SAIH can progressively evaluate the AI performance trend of HPC systems, e.g., increasing from 5.2% to 59.6% of the peak theoretical hardware performance. The evaluation results are analyzed and summarized into insight findings on performance issues. For instance, we find that the AI application constantly consumes the I/O bandwidth of the shared parallel file system during its iteratively training model. If I/O contention exists, the shared parallel file system might become a bottleneck.

We present SinGRAV, an attempt to learn a generative radiance volume from multi-view observations of a single natural scene, in stark contrast to existing category-level 3D generative models that learn from images of many object-centric scenes. Inspired by SinGAN, we also learn the internal distribution of the input scene, which necessitates our key designs w.r.t. the scene representation and network architecture. Unlike popular multi-layer perceptrons (MLP)-based architectures, we particularly employ convolutional generators and discriminators, which inherently possess spatial locality bias, to operate over voxelized volumes for learning the internal distribution over a plethora of overlapping regions. On the other hand, localizing the adversarial generators and discriminators over confined areas with limited receptive fields easily leads to highly implausible geometric structures in the spatial. Our remedy is to use spatial inductive bias and joint discrimination on geometric clues in the form of 2D depth maps. This strategy is effective in improving spatial arrangement while incurring negligible additional computational cost. Experimental results demonstrate the ability of SinGRAV in generating plausible and diverse variations from a single scene, the merits of SinGRAV over state-of-the-art generative neural scene models, and the versatility of SinGRAV by its use in a variety of applications. Code and data will be released to facilitate further research.

Traditional Chinese painting (TCP) is an invaluable cultural heritage resource and a unique visual art style. In recent years, there has been a growing emphasis on the digitalization of TCP for cultural preservation and revitalization. The resulting digital copies have enabled the advancement of computational methods for a structured and systematic understanding of TCP. To explore this topic, we conduct an in-depth analysis of 94 pieces of literature. We examine the current use of computer technologies on TCP from three perspectives, based on numerous conversations with specialists. First, in light of the “Six Principles of Painting” theory, we categorize the articles according to their research focus on artistic elements. Second, we create a four-stage framework to illustrate the purposes of TCP applications. Third, we summarize the popular computational techniques applied to TCP. This work also provides insights into potential applications and prospects, with professional opinion.

As a highly vascular eye part, the choroid is crucial in various eye disease diagnoses. However, limited research has focused on the inner structure of the choroid due to the challenges in obtaining sufficient accurate label data, particularly for the choroidal vessels. Meanwhile, the existing direct choroidal vessel segmentation methods for the intelligent diagnosis of vascular assisted ophthalmic diseases are still unsatisfactory due to noise data, while the synergistic segmentation methods compromise vessel segmentation performance for the choroid layer segmentation tasks. Common cascaded structures grapple with error propagation during training. To address these challenges, we propose a cascade learning segmentation method for the inner vessel structures of the choroid in this paper. Specifically, we propose Transformer-Assisted Cascade Learning Network (TACLNet) for choroidal vessel segmentation, which comprises a two-stage training strategy: pre-training for choroid layer segmentation and joint training for choroid layer and choroidal vessel segmentation. We also enhance the skip connection structures by introducing a multi-scale subtraction connection module designated as MSC, capturing differential and detailed information simultaneously. Additionally, we implement an auxiliary Transformer branch named ATB to integrate global features into the segmentation process. Experimental results exhibit that our method achieves the state-of-the-art performance for choroidal vessel segmentation. Besides, we further validate the significant superiority of the proposed method for retinal fluid segmentation in optical coherence tomography (OCT) scans on a publicly available dataset. All these fully prove that our TACLNet contributes to the advancement of choroidal vessel segmentation and is of great significance for ophthalmic research and clinical application.

Mobile applications (apps for short) often need to display images. However, inefficient image displaying (IID) issues are pervasive in mobile apps, and can severely impact app performance and user experience. This paper first establishes a descriptive framework for the image displaying procedures of IID issues. Based on the descriptive framework, we conduct an empirical study of 216 real-world IID issues collected from 243 popular open-source Android apps to validate the presence and severity of IID issues, and then shed light on these issues’ characteristics to support research on effective issue detection. With the findings of this study, we propose a static IID issue detection tool TAPIR and evaluate it with 243 real-world Android apps. Encouragingly, 49 and 64 previously-unknown IID issues in two different versions of 16 apps reported by TAPIR are manually confirmed as true positives, respectively, and 16 previously-unknown IID issues reported by TAPIR have been confirmed by developers and 13 have been fixed. Then, we further evaluate the performance impact of these detected IID issues and the performance improvement if they are fixed. The results demonstrate that the IID issues detected by TAPIR indeed cause significant performance degradation, which further show the effectiveness and efficiency of TAPIR.

While databases are widely-used in commercial user-facing services that have stringent quality-of-service (QoS) requirement, it is crucial to ensure their good performance and minimize the hardware usage at the same time. Our investigation shows that the optimal DBMS (database management system) software configuration varies for different user request patterns (i.e., workloads) and hardware configurations. It is challenging to identify the optimal software and hardware configurations for a database workload, because DBMSs have hundreds of tunable knobs, the effect of tuning a knob depends on other knobs, and the dependency relationship changes under different hardware configurations. In this paper, we propose SHA, a software and hardware auto-tuning system for DBMSs. SHA is comprised of a scaling-based performance predictor, a reinforcement learning (RL) based software tuner, and a QoS-aware resource reallocator. The performance predictor predicts its optimal performance with different hardware configurations and identifies the minimum amount of resources for satisfying its performance requirement. The software tuner fine-tunes the DBMS software knobs to optimize the performance of the workload. The resource reallocator assigns the saved resources to other applications to improve resource utilization without incurring QoS violation of the database workload. Experimental results show that SHA improves the performance of database workloads by 9.9% on average compared with a state-of-the-art solution when the hardware configuration is fixed, and improves 43.2% of resource utilization while ensuring the QoS.