Master Thesis: Uncertainty Estimation for Segmentation in Autonomous Driving
In the context of Autonomous Driving, it is crucial to have a measure of the uncertainty associated to the various predictions performed by the Deep Learning models. This helps not only to combine various predictions from different pipelines, but also to understand the real confidence associated to each prediction. Networks tend to be overly confident (~99%, as derived from the logits probabilities) also on wrong predictions, or rather unknown data and scenarios. This makes such confidence unreliable. Therefore, uncertainty estimation complements predictions by quantifying how certain the models really are, with respect to the inputs, or their own weights and the way they were trained.
Among the fundamental tasks of Autonomous Driving there is segmentation (semantic segmentation, instance segmentation, panoptic segmentation, part segmentation...), which is where we would like to integrate uncertainty estimation with this work.
This Master Thesis will be done in cooperation with BMW.
Example of uncertainty estimation for semantic segmentation.
source: Postels, Janis, Francesco Ferroni, Huseyin Coskun, Nassir Navab, and Federico Tombari. "Sampling-free Epistemic Uncertainty Estimation Using Approximated Variance Propagation" - ICCV 2019
• You will start from state-of-the-art Deep Learning techniques and prior work from our Team, to implement an uncertainty estimation method for segmentation tasks in Autonomous Driving.
• You will prepare datasets and run various experiments to evaluate your findings.
• You may develop a novel uncertainty estimation method, which might lead to a publication at an international peer-reviewed conference or journal, contributing to the scientific community.
• You will work at BMW in an international and diverse team of developers and students at the Autonomous Driving Campus.
• You are familiar with Deep Learning and have a good understanding of Machine Learning and Computer Vision.
• You know how to train a Convolutional Neural Network.
• You have solid skills in PyTorch?
and Python, and familiarity with Linux OS.
Postels, Janis, Francesco Ferroni, Huseyin Coskun, Nassir Navab, and Federico Tombari. "Sampling-free Epistemic Uncertainty Estimation Using Approximated Variance Propagation.", in ICCV 2019.
Lee, Hong Joo, Seong Tae Kim, Nassir Navab, and Yong Man Ro. "Efficient Ensemble Model Generation for Uncertainty Estimation with Bayesian Approximation in Segmentation.", in arXiv 2020.
If you are interested, please contact Stefano at email@example.com
with your latest transcript and CV for inquiries and applications.