Semantic segmentation with inexpensive simulated data

Jukka Peltomäki; Mengyang Chen; Heikki Huttunen

doi:10.1109/NORCHIP.2019.8906954

Semantic segmentation with inexpensive simulated data

Jukka Peltomäki, Mengyang Chen, Heikki Huttunen

Computing Sciences

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

Abstract

This paper studies the benefits of adding inexpensively gathered simulated data to improve the training of semantic segmentation models. We introduce our implementation to gather simulated datasets with minimal effort. In our implementation, we utilize a commonly available game engine (Unity) and aux-illiary graphical assets to assemble an environment to generate simulated data inexpensively. We also demonstrate that even the usage of spartan simulated data mixed with real-life data can increase the performance of the trained model slightly, given that the ratio of simulated data is suitable for the datasets.

Original language	English
Title of host publication	2019 IEEE Nordic Circuits and Systems Conference (NORCAS): NORCHIP and International Symposium of System-on-Chip (SoC)
Publisher	IEEE
Number of pages	6
ISBN (Electronic)	978-1-7281-2769-9
ISBN (Print)	978-1-7281-2770-5
DOIs	https://doi.org/10.1109/NORCHIP.2019.8906954
Publication status	Published - Oct 2019
Publication type	A4 Article in conference proceedings
Event	IEEE Nordic Circuits and Systems Conference - Duration: 1 Jan 2000 → …

Conference

Conference	IEEE Nordic Circuits and Systems Conference
Period	1/01/00 → …

Publication forum classification

Publication forum level 1

Access to Document

10.1109/NORCHIP.2019.8906954

Cite this

@inproceedings{1e4038c70c85428db2e2eb9f025e41f2,

title = "Semantic segmentation with inexpensive simulated data",

abstract = "This paper studies the benefits of adding inexpensively gathered simulated data to improve the training of semantic segmentation models. We introduce our implementation to gather simulated datasets with minimal effort. In our implementation, we utilize a commonly available game engine (Unity) and aux-illiary graphical assets to assemble an environment to generate simulated data inexpensively. We also demonstrate that even the usage of spartan simulated data mixed with real-life data can increase the performance of the trained model slightly, given that the ratio of simulated data is suitable for the datasets.",

author = "Jukka Peltom{\"a}ki and Mengyang Chen and Heikki Huttunen",

note = "INT=COMP, {"}Chen, Mengyang{"}; IEEE Nordic Circuits and Systems Conference ; Conference date: 01-01-2000",

year = "2019",

month = oct,

doi = "10.1109/NORCHIP.2019.8906954",

language = "English",

isbn = "978-1-7281-2770-5",

booktitle = "2019 IEEE Nordic Circuits and Systems Conference (NORCAS): NORCHIP and International Symposium of System-on-Chip (SoC)",

publisher = "IEEE",

}

TY - GEN

T1 - Semantic segmentation with inexpensive simulated data

AU - Peltomäki, Jukka

AU - Chen, Mengyang

AU - Huttunen, Heikki

N1 - INT=COMP, "Chen, Mengyang"

PY - 2019/10

Y1 - 2019/10

N2 - This paper studies the benefits of adding inexpensively gathered simulated data to improve the training of semantic segmentation models. We introduce our implementation to gather simulated datasets with minimal effort. In our implementation, we utilize a commonly available game engine (Unity) and aux-illiary graphical assets to assemble an environment to generate simulated data inexpensively. We also demonstrate that even the usage of spartan simulated data mixed with real-life data can increase the performance of the trained model slightly, given that the ratio of simulated data is suitable for the datasets.

AB - This paper studies the benefits of adding inexpensively gathered simulated data to improve the training of semantic segmentation models. We introduce our implementation to gather simulated datasets with minimal effort. In our implementation, we utilize a commonly available game engine (Unity) and aux-illiary graphical assets to assemble an environment to generate simulated data inexpensively. We also demonstrate that even the usage of spartan simulated data mixed with real-life data can increase the performance of the trained model slightly, given that the ratio of simulated data is suitable for the datasets.

U2 - 10.1109/NORCHIP.2019.8906954

DO - 10.1109/NORCHIP.2019.8906954

M3 - Conference contribution

SN - 978-1-7281-2770-5

BT - 2019 IEEE Nordic Circuits and Systems Conference (NORCAS): NORCHIP and International Symposium of System-on-Chip (SoC)

PB - IEEE

T2 - IEEE Nordic Circuits and Systems Conference

Y2 - 1 January 2000

ER -

Semantic segmentation with inexpensive simulated data

Abstract

Conference

Publication forum classification

Access to Document

Fingerprint

Cite this