Home > Research > Publications & Outputs > Semantic Segmentation under Adverse Conditions

Research

Electronic data

Keywords

View graph of relations

Semantic Segmentation under Adverse Conditions: A Weather and Nighttime-aware Synthetic Data-based Approach

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper › peer-review

Forthcoming

Standard

Semantic Segmentation under Adverse Conditions: A Weather and Nighttime-aware Synthetic Data-based Approach. / Kerim, Abdulrahman; Chamone, Felipe; De Souza Ramos, Washington et al.
The British Machine Vision Conference (BMVC). 2022.

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSN › Conference contribution/Paper › peer-review

Harvard

Kerim, A, Chamone, F, De Souza Ramos, W, Soriano Marcolino, L, Nascimento, ER & Jiang, R 2022, Semantic Segmentation under Adverse Conditions: A Weather and Nighttime-aware Synthetic Data-based Approach. in The British Machine Vision Conference (BMVC).

APA

Kerim, A., Chamone, F., De Souza Ramos, W., Soriano Marcolino, L., Nascimento, E. R., & Jiang, R. (in press). Semantic Segmentation under Adverse Conditions: A Weather and Nighttime-aware Synthetic Data-based Approach. In The British Machine Vision Conference (BMVC)

Vancouver

Kerim A, Chamone F, De Souza Ramos W, Soriano Marcolino L, Nascimento ER, Jiang R. Semantic Segmentation under Adverse Conditions: A Weather and Nighttime-aware Synthetic Data-based Approach. In The British Machine Vision Conference (BMVC). 2022

Author

Kerim, Abdulrahman ; Chamone, Felipe ; De Souza Ramos, Washington et al. / Semantic Segmentation under Adverse Conditions : A Weather and Nighttime-aware Synthetic Data-based Approach. The British Machine Vision Conference (BMVC). 2022.

Bibtex

@inproceedings{de67475db3474081bcde536542b8e6dc,
title = "Semantic Segmentation under Adverse Conditions: A Weather and Nighttime-aware Synthetic Data-based Approach",
abstract = "Recent semantic segmentation models perform well under standard weather conditions and sufficient illumination but struggle with adverse weather conditions and nighttime. Collecting and annotating training data under these conditions is expensive, timeconsuming, error-prone, and not always practical. Usually, synthetic data is used as a feasible data source to increase the amount of training data. However, just directly using synthetic data may actually harm the model{\textquoteright}s performance under normal weather conditions while getting only small gains in adverse situations. Therefore, we present a novel architecture specifically designed for using synthetic training data for domain adaptation. We propose a simple yet powerful addition to DeepLabV3+ by using weather and time-of-the-day supervisors trained with multi-task learning, making it both weatherand nighttime aware, which improves its mIoU accuracy by 14 percentage points on the ACDC dataset while maintaining a score of 75% mIoU on the Cityscapes dataset. Our code is available at https://github.com/lsmcolab/Semantic-Segmentation-under-Adverse-Conditions.",
keywords = "Semantic Segmentation, Domain Adaptation, Synthetic Data, Adverse Conditions",
author = "Abdulrahman Kerim and Felipe Chamone and {De Souza Ramos}, Washington and {Soriano Marcolino}, Leandro and Nascimento, {Erickson R.} and Richard Jiang",
year = "2022",
month = sep,
day = "30",
language = "English",
booktitle = "The British Machine Vision Conference (BMVC)",

}

RIS

TY - GEN

T1 - Semantic Segmentation under Adverse Conditions

T2 - A Weather and Nighttime-aware Synthetic Data-based Approach

AU - Kerim, Abdulrahman

AU - Chamone, Felipe

AU - De Souza Ramos, Washington

AU - Soriano Marcolino, Leandro

AU - Nascimento, Erickson R.

AU - Jiang, Richard

PY - 2022/9/30

Y1 - 2022/9/30

N2 - Recent semantic segmentation models perform well under standard weather conditions and sufficient illumination but struggle with adverse weather conditions and nighttime. Collecting and annotating training data under these conditions is expensive, timeconsuming, error-prone, and not always practical. Usually, synthetic data is used as a feasible data source to increase the amount of training data. However, just directly using synthetic data may actually harm the model’s performance under normal weather conditions while getting only small gains in adverse situations. Therefore, we present a novel architecture specifically designed for using synthetic training data for domain adaptation. We propose a simple yet powerful addition to DeepLabV3+ by using weather and time-of-the-day supervisors trained with multi-task learning, making it both weatherand nighttime aware, which improves its mIoU accuracy by 14 percentage points on the ACDC dataset while maintaining a score of 75% mIoU on the Cityscapes dataset. Our code is available at https://github.com/lsmcolab/Semantic-Segmentation-under-Adverse-Conditions.

AB - Recent semantic segmentation models perform well under standard weather conditions and sufficient illumination but struggle with adverse weather conditions and nighttime. Collecting and annotating training data under these conditions is expensive, timeconsuming, error-prone, and not always practical. Usually, synthetic data is used as a feasible data source to increase the amount of training data. However, just directly using synthetic data may actually harm the model’s performance under normal weather conditions while getting only small gains in adverse situations. Therefore, we present a novel architecture specifically designed for using synthetic training data for domain adaptation. We propose a simple yet powerful addition to DeepLabV3+ by using weather and time-of-the-day supervisors trained with multi-task learning, making it both weatherand nighttime aware, which improves its mIoU accuracy by 14 percentage points on the ACDC dataset while maintaining a score of 75% mIoU on the Cityscapes dataset. Our code is available at https://github.com/lsmcolab/Semantic-Segmentation-under-Adverse-Conditions.

KW - Semantic Segmentation

KW - Domain Adaptation

KW - Synthetic Data

KW - Adverse Conditions

M3 - Conference contribution/Paper

BT - The British Machine Vision Conference (BMVC)

ER -