Data Augmentation Methods for Semantic Segmentation-based Mobile Robot Perception System
Апстракт
Data augmentation has become a standard technique for increasing deep learning models’ accuracy and robustness. Different pixel intensity modifications, image transformations, and noise additions represent the most utilized data augmentation methods. In this paper, a comprehensive evaluation of data augmentation techniques for mobile robot perception system is performed. The perception system based on a deep learning model for semantic segmentation is augmented by 17 techniques to obtain better generalization characteristics during the training process. The deep learning model is trained and tested on a custom dataset and utilized in real-time scenarios. The experimental results show the increment of 6.2 in mIoU (mean Intersection over Union) for the best combination of data augmentation strategies.
Кључне речи:
Mobile robot perception system / Deep learning / Data augmentation / Semantic segmentationИзвор:
Serbian Journal of Electrical Engineering, 2022, 19, 3, 291-302-Финансирање / пројекти:
- Министарство науке, технолошког развоја и иновација Републике Србије, институционално финансирање - 200105 (Универзитет у Београду, Машински факултет) (RS-MESTD-inst-2020-200105)
- MISSION4.0 - Deep Machine Learning and Swarm Intelligence-Based Optimization Algorithms for Control and Scheduling of Cyber-Physical Systems in Industry 4.0 (RS-ScienceFundRS-AI-6523109)
Колекције
Институција/група
Mašinski fakultetTY - JOUR AU - Jokić, Aleksandar AU - Đokić, Lazar AU - Petrović, Milica AU - Miljković, Zoran PY - 2022 UR - https://machinery.mas.bg.ac.rs/handle/123456789/3967 AB - Data augmentation has become a standard technique for increasing deep learning models’ accuracy and robustness. Different pixel intensity modifications, image transformations, and noise additions represent the most utilized data augmentation methods. In this paper, a comprehensive evaluation of data augmentation techniques for mobile robot perception system is performed. The perception system based on a deep learning model for semantic segmentation is augmented by 17 techniques to obtain better generalization characteristics during the training process. The deep learning model is trained and tested on a custom dataset and utilized in real-time scenarios. The experimental results show the increment of 6.2 in mIoU (mean Intersection over Union) for the best combination of data augmentation strategies. T2 - Serbian Journal of Electrical Engineering T1 - Data Augmentation Methods for Semantic Segmentation-based Mobile Robot Perception System IS - 3 SP - 291-302 VL - 19 DO - https://doi.org/10.2298/SJEE2203291J ER -
@article{ author = "Jokić, Aleksandar and Đokić, Lazar and Petrović, Milica and Miljković, Zoran", year = "2022", abstract = "Data augmentation has become a standard technique for increasing deep learning models’ accuracy and robustness. Different pixel intensity modifications, image transformations, and noise additions represent the most utilized data augmentation methods. In this paper, a comprehensive evaluation of data augmentation techniques for mobile robot perception system is performed. The perception system based on a deep learning model for semantic segmentation is augmented by 17 techniques to obtain better generalization characteristics during the training process. The deep learning model is trained and tested on a custom dataset and utilized in real-time scenarios. The experimental results show the increment of 6.2 in mIoU (mean Intersection over Union) for the best combination of data augmentation strategies.", journal = "Serbian Journal of Electrical Engineering", title = "Data Augmentation Methods for Semantic Segmentation-based Mobile Robot Perception System", number = "3", pages = "291-302", volume = "19", doi = "https://doi.org/10.2298/SJEE2203291J" }
Jokić, A., Đokić, L., Petrović, M.,& Miljković, Z.. (2022). Data Augmentation Methods for Semantic Segmentation-based Mobile Robot Perception System. in Serbian Journal of Electrical Engineering, 19(3), 291-302. https://doi.org/https://doi.org/10.2298/SJEE2203291J
Jokić A, Đokić L, Petrović M, Miljković Z. Data Augmentation Methods for Semantic Segmentation-based Mobile Robot Perception System. in Serbian Journal of Electrical Engineering. 2022;19(3):291-302. doi:https://doi.org/10.2298/SJEE2203291J .
Jokić, Aleksandar, Đokić, Lazar, Petrović, Milica, Miljković, Zoran, "Data Augmentation Methods for Semantic Segmentation-based Mobile Robot Perception System" in Serbian Journal of Electrical Engineering, 19, no. 3 (2022):291-302, https://doi.org/https://doi.org/10.2298/SJEE2203291J . .