TY - GEN
T1 - Trade-off analysis between parallelism and accuracy of SLIC on apache spark
AU - Park, Gang Min
AU - Heo, Yong Seok
AU - Kwon, Hyuk Yoon
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021/1
Y1 - 2021/1
N2 - In this paper, we present a parallel algorithm for SLIC on Apache Spark, which we call PSLIC-on-Spark. To this purpose, we have extended the original SLIC algorithm to use the operations in Apache Spark, supporting its parallel processing on multiple executors in the Apache Spark cluster. Then, we analyze the trade-off relationship of PSLIC-on-Spark between its processing speed and accuracy due to partitioning of the original image data sets. Especially, we identify two limitations in PSLIC-on-Spark, which degrade the accuracy of the original SLIC. Through experiments, we verify the trade-off relationship. Specifically, we show that PSLIC-on-Spark using 8 CPU cores reduces the processing time of SLIC by 2. 24∼2.93 times while it reduces the boundary recall (BR) of SLIC by 1. 54∼6.32 % and increases under-segmentation error (UE) by 1. 79∼6.2 %. In contrast, PSLIC-on-Spark using 2 CPU cores reduces the processing time of SLIC by 1.38∼1.45 times while it reduces the BR of SLIC by 0. 28∼1.5 %, and increases UE by 0. 25∼1.77 %. We also verify the effectiveness of PSLIC-on-Spark to deal with a large-scale image by showing that the processing speed of PSLIC-on-Spark becomes much more efficient as the image size becomes large. Specifically, compared to the original SLIC, the proposed SLIC-on-Spark reduces its processing time by 2.23 times for the image of 480×320 pixels and by 5.59
AB - In this paper, we present a parallel algorithm for SLIC on Apache Spark, which we call PSLIC-on-Spark. To this purpose, we have extended the original SLIC algorithm to use the operations in Apache Spark, supporting its parallel processing on multiple executors in the Apache Spark cluster. Then, we analyze the trade-off relationship of PSLIC-on-Spark between its processing speed and accuracy due to partitioning of the original image data sets. Especially, we identify two limitations in PSLIC-on-Spark, which degrade the accuracy of the original SLIC. Through experiments, we verify the trade-off relationship. Specifically, we show that PSLIC-on-Spark using 8 CPU cores reduces the processing time of SLIC by 2. 24∼2.93 times while it reduces the boundary recall (BR) of SLIC by 1. 54∼6.32 % and increases under-segmentation error (UE) by 1. 79∼6.2 %. In contrast, PSLIC-on-Spark using 2 CPU cores reduces the processing time of SLIC by 1.38∼1.45 times while it reduces the BR of SLIC by 0. 28∼1.5 %, and increases UE by 0. 25∼1.77 %. We also verify the effectiveness of PSLIC-on-Spark to deal with a large-scale image by showing that the processing speed of PSLIC-on-Spark becomes much more efficient as the image size becomes large. Specifically, compared to the original SLIC, the proposed SLIC-on-Spark reduces its processing time by 2.23 times for the image of 480×320 pixels and by 5.59
KW - Accuracy
KW - Apache Spark
KW - Image Segmentation
KW - Parallel Processing
KW - SLIC
UR - https://www.scopus.com/pages/publications/85102977084
U2 - 10.1109/BigComp51126.2021.00011
DO - 10.1109/BigComp51126.2021.00011
M3 - Conference contribution
AN - SCOPUS:85102977084
T3 - Proceedings - 2021 IEEE International Conference on Big Data and Smart Computing, BigComp 2021
SP - 5
EP - 12
BT - Proceedings - 2021 IEEE International Conference on Big Data and Smart Computing, BigComp 2021
A2 - Unger, Herwig
A2 - Kim, Jinho
A2 - Kang, U
A2 - So-In, Chakchai
A2 - Du, Junping
A2 - Saad, Walid
A2 - Ha, Young-guk
A2 - Wagner, Christian
A2 - Bourgeois, Julien
A2 - Sathitwiriyawong, Chanboon
A2 - Kwon, Hyuk-Yoon
A2 - Leung, Carson
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2021 IEEE International Conference on Big Data and Smart Computing, BigComp 2021
Y2 - 17 January 2021 through 20 January 2021
ER -