Implementation of the parallel mean shift-based image segmentation algorithm on a GPU cluster

Huang, Fang; Chen, Yinjie; Li, Li; Zhou, Ji; Jian, Tao; Tan, Xicheng; Fan, Guangsong

doi:10.1080/17538947.2018.1432709

Cited by 21 publications

(24 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The simplest form of clustering is called partitional clustering, which is mainly to divide a given data set into disjoint clusters. In the partitional clustering problem, each cluster with approximate similar points is a very common scenario, such as image segmentation [8] and object tracking [9] and movement detection [10], [11]. In order to solve this problem, many clustering algorithms have been proposed.…”

Section: Introductionmentioning

confidence: 99%

“…Among the unsupervised clustering algorithms, mean-shift [13] is arguably the one of the most widely used clustering algorithm in clustering problems, which has been used in image segmentation [8], voice processing [14], [15], object tracking [9] and vector embedding machine learning [16]. The advantage of mean-shift is a density-based(centroid-based) clustering approach and can determine the number of clusters adaptively.…”

Section: Introductionmentioning

confidence: 99%

“…For example, the fast mean-shift algorithm [21] was developed to achieve significant speed-up compared with CPU based mean-shift clustering. Recently, further accelerated computational speed with parallel tensor operations has been achieved [8]. However, the algorithm is memory extensive, which is infeasible for processing large data-set (e.g., 512×512 and 1024×1024 pixel points in Figure 1) using the ordinary GPU cards.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

GPU-accelerated Faster Mean Shift with euclidean distance metrics

You¹,

Han²,

Hu³

et al. 2021

Preprint

View full text Add to dashboard Cite

Handling clustering problems are important in data statistics, pattern recognition and image processing. The mean-shift algorithm, a common unsupervised algorithms, is widely used to solve clustering problems. However, the meanshift algorithm is restricted by its huge computational resource cost. In previous research [1], we proposed a novel GPUaccelerated Faster Mean-shift algorithm, which greatly speed up the cosine-embedding clustering problem. In this study, we extend and improve the previous algorithm to handle Euclidean distance metrics. Different from conventional GPU-based meanshift algorithms, our algorithm adopts novel Seed Selection & Early Stopping approaches, which greatly increase computing speed and reduce GPU memory consumption. In the simulation testing, when processing a 200K points clustering problem, our algorithm achieved around 3 times speedup compared to the state-of-the-art GPU-based mean-shift algorithms with optimized GPU memory consumption. Moreover, in this study, we implemented a plug-and-play model for faster mean-shift algorithm, which can be easily deployed.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

GPU-accelerated Faster Mean Shift with euclidean distance metrics

You¹,

Han²,

Hu³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…are all challenging [31]. Among these, to explore and exploit multi-level parallelism of applications on a GPU cluster is of great importance to solve the efficiency problem, however, very few work has addressed this issue in the RS community [32].…”

Section: Introductionmentioning

confidence: 99%

High-Performance Time-Series Quantitative Retrieval From Satellite Images on a GPU Cluster

Liu

Xue

Ren

et al. 2019

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

The quality and accuracy of remote sensing instruments continue to increase, allowing geoscientists to perform various quantitative retrieval applications to observe the geophysical variables of land, atmosphere, ocean etc. The explosive growth of time-series RS data over large-scales poses great challenges on managing, processing and interpreting RS ''Big Data''. To explore these time series RS data efficiently, in this paper, we design and implement a high performance framework to address the time-consuming time-series quantitative retrieval issue on a GPU cluster, taking the aerosol optical depth (AOD) retrieval from satellite images as a study case. The presented framework exploits the multi-level parallelism for time-series quantitative RS retrieval to promote efficiency. At the coarse-grained level of parallelism, the AOD time series retrieval is represented as multi-directed acyclic graph (DAG) workflows and scheduled based on a list-based heuristic algorithm, heterogeneous earliest finish time (HEFT), taking the idle slot and priorities of retrieval jobs into account. At the fine-grained level, the parallel strategies for the major remote sensing image processing algorithms divided into three categories, i.e. the point or pixel-based operations, the local operations and the global or irregular operations have been summarized. The parallel framework was implemented with message passing interface (MPI) and compute unified device architecture (CUDA), and experimental results with the AOD retrieval case verify the effectiveness of presented framework.

show abstract

“…Recently, various EO data processing tasks, which are computationally expensive in nature, have been accelerated using HPC approaches such as CPU+GPU hybrid cluster computing and cloud computing. The work by Huang et al developed a parallel mean shift segmentation algorithm to deal with the large number of RS images in real‐world applications and tested the applicability of the proposed approach on Shelob, which is a CPU+GPU heterogeneous HPC platform at Louisiana State University. Shelob contains 32 compute nodes, with each node consisting of one 16‐core CPU and two NVIDIA K20 GPUs.…”

Section: Introductionmentioning

confidence: 99%

Performance analysis and optimization for scalable deployment of deep learning models for country‐scale settlement mapping on Titan supercomputer

Kurte

Sanyal

Berres

et al. 2019

Concurrency and Computation

View full text Add to dashboard Cite

This paper presents a scalable object detection workflow for detecting objects, such as settlements, from remotely sensed (RS) imagery. We have successfully deployed this workflow on Titan supercomputer and utilized it for the task of mapping human settlement at a country scale.The performance of various stages in the workflow was analyzed before making it operational.The workflow implemented various strategies to address issues such as suboptimal resource utilization and long-tail effects due to unbalanced image workload, data loss due to runtime failures, and maximum wall-time constraints imposed by Titan's job scheduling policy. A mean shift clustering-based static load balancing strategy was implemented, which partitions the image load such that each partition contained similar-sized images. Furthermore, a checkpoint-restart strategy was added in the workflow as a fault-tolerance mechanism to prevent the data losses due to unforeseen runtime failures. The performance of the above-mentioned strategies was observed in various scenarios, such as node failure, exceeding wall time, and successful completion. Using this workflow, we have processed an RS data set that has a spatial resolution of 0.31 m and is comprised of 685 675 km 2 of area of the Republic of Zambia in under six hours using 5426 nodes of the Titan supercomputer. KEYWORDS convolutional neural network, deep learning, fault tolerance, HPC, human settlement mapping, load balancing 1 INTRODUCTION The recent development in sensor electronics is enabling remote sensing (RS) satellites to capture data at a submeter scale. This advancement in RS technology offers new opportunities to understand Earth's features at a finer spatial scale and motivates the development of novel approaches for complex problems such as mapping human settlements. However, as an effect of the development in sensor technology, a rapid growth in the cumulative volume of Earth observation (EO) archives has been observed during the last few years. 1 A typical very-high-resolution RS imagery contains billions of pixels, and for a country-scale mapping, thousands of such images need to be processed, which is computationally demanding. This computational load can pose problems during time-critical events, such as with flooding and wild fire natural disasters, when real-time/near-real-time mapping of human settlement in the affected area is necessary. However, high-performance computing (HPC)-driven approaches for processing such huge volume of data can mitigate this problem.

show abstract

Implementation of the parallel mean shift-based image segmentation algorithm on a GPU cluster

Cited by 21 publications

References 27 publications

GPU-accelerated Faster Mean Shift with euclidean distance metrics

GPU-accelerated Faster Mean Shift with euclidean distance metrics

High-Performance Time-Series Quantitative Retrieval From Satellite Images on a GPU Cluster

Performance analysis and optimization for scalable deployment of deep learning models for country‐scale settlement mapping on Titan supercomputer

Contact Info

Product

Resources

About