Modeling and Learning from Sensor Effects
DESCRIPTION
Varying sensor effects can degrade performance and generalizability of results for visual tasks trained on human annotated datasets.We developed efficient, automated physically-based augmentation pipelines to vary sensor effects – specifically, chromatic aberration, blur, exposure, noise, and color cast – across both real and synthetic imagery. We also developed an automated method to identify mistakes made by object detectors without ground truth labels, based on temporal and stereo inconsistencies. Additionally, we are interested in exploring alternative sensor modalities.
Related Publications
A. Carlson, R. Vasudevan, and M. Johnson-Roberson. "Shadow Transfer: Single Image Relighting For Urban Road Scenes." arXiv preprint arXiv:1909.10363 (2019). [PDF]
A. Carlson, et al. "Sensor Transfer: Learning Optimal Sensor Effect Image Augmentation for Sim-to-Real Domain Adaptation." IEEE Robotics and Automation Letters 4.3 (2019): 2431-2438. [PDF]
M. S. Ramanagopal, et al. "Failing to learn: Autonomously identifying perception failures for self-driving cars." IEEE Robotics and Automation Letters 3.4 (2018): 3860-3867. [PDF][Website]
@article{carlson2019shadow,
title={Shadow Transfer: Single Image Relighting For Urban Road Scenes},
author={Carlson, Alexandra and Vasudevan, Ram and Johnson-Roberson, Matthew},
journal={arXiv preprint arXiv:1909.10363},
year={2019}
}
@article{carlson2019sensor,
title={Sensor Transfer: Learning Optimal Sensor Effect Image Augmentation for Sim-to-Real Domain Adaptation},
author={Carlson, Alexandra and Skinner, Katherine A and Vasudevan, Ram and Johnson-Roberson, Matthew},
journal={IEEE Robotics and Automation Letters},
volume={4},
number={3},
pages={2431--2438},
year={2019},
publisher={IEEE}
}
@article{ramanagopal2018failing,
title={Failing to learn: Autonomously identifying perception failures for self-driving cars},
author={Ramanagopal, Manikandasriram Srinivasan and Anderson, Cyrus and Vasudevan, Ram and Johnson-Roberson, Matthew},
journal={IEEE Robotics and Automation Letters},
volume={3},
number={4},
pages={3860--3867},
year={2018},
publisher={IEEE}
}