Robust Understanding of Street Scenes Using Computer Vision

Summer School and Workshop on Uncertainty in Computer Vision for Automated Driving

Deep neural networks lead to stunning performance in semantic interpretation of street scenes. Nevertheless, a number of safety relevant questions still remain under intensive research. In particular this applies to the question, whether neural networks can successfully self-monitor their performance by quantifying the uncertainty of predictions, and which kind of data they need to learn this. In this workshop, we bring together experts on uncertainty in machine learning and computer vision to present state of the art research towards enhancing the safety of automated driving.

Figure: Visualization of Exemplary Methods for Prediction Uncertainty Quantification in Street Scenes

Relevant topics of interest for this workshop include (but are not limited to) the following:

  • Quantification of uncertainty
  • Detection of out-of-distribution samples
  • Resilience to adversarial attacks
  • Open world recognition
  • Domain shift detection
  • Domain adaptation
  • Cross-domain learning and robustness to real-world conditions
  • Sensor fusion and other types of redundancy


The lectures, presentations and discussions will take place at:
University of Zagreb
Grey Hall at the Faculty of Electrical Engineering and Computing (FER)
Unska ul. 3, 10000, Zagreb, Croatia.

Show in browser.

Schedule Summer School

Wednesday, 09-21-2022

Hanno Gottschalk, University of Wuppertal
From Information to Outliers
10:35-11:00 Coffee Break
11:00-12:30 Matthias Rottmann, EPFL
Uncertainty in Machine Learning -- Notions of Uncertainty
12:30-13:45 Lunch break
13:45-15:15 Matthias Rottmann, EPFL
Uncertainty in Machine Learning -- Methods for Uncertainty Quantification

Schedule Workshop

Thursday, 09-22-2022

How Synthetic Training Powers Anomaly and Obstacle Detection in Traffic Scenes
09:45-10:30 , ETH Zurich
Leveraging Multi-Modality for Robust Scene Understanding
10:30-10:40 Coffee Break
10:40-11:00 , CARIAD
Enabling AI in Safety-Critical Automotive Products
11:00-11:20 , Aptiv
Automotive Big Data Video Analysis with AiBox
11:20-11:40 , dSPACE
Challenges in Data Selection: How to Fish in the Data Lake
11:40-12:00 , Intel
Learning Confidence Classification from Richly Annotated Synthetic Data
Poster Session
13:00-14:00 Lunch break
14:00-14:30 , University of Wuppertal
Tracking and Retrieval of Out of Distribution Objects in Video Sequences
14:30-15:00 , Bielefeld University
SegmentMeIfYouCan: A Benchmark for Anomaly Segmentation
15:00-15:30 , University of Wuppertal
Gradient-Based Quantification of Epistemic Uncertainty for Deep Object Detectors
15:30-15:45 Coffee Break
15:45-16:15 , Ruhr University Bochum
Improving Robustness under Domain Shift of Semantic Segmentation by Depth Estimation
16:15-16:45 , University of Wuppertal
Domain Adaptation with Generative Adversarial Networks
16:45-17:05 , Rimac Automobili
Challenges in Scene Understanding for Autonomous Racing

Friday, 09-23-2022

09:00-09:45 , AIT Vienna
Robust Evaluation of Computer Vision for Autonomous Driving
09:45-10:30 , ETH Zurich
Creating Synthetic and Real Data for Semantic Scene Understanding in Adverse Conditions
10:30-10:45 Coffee Break
10:45-11:30 , EPFL
Automated Detection of Labeling Errors in Semantic Segmentation Datasets
11:30-12:15 , Czech Technical University in Prague
Road Anomaly Detection by Generative and Discriminative Road Appearance Modelling
12:15-14:00 Lunch break
14:00-14:20 , Gideon Brothers
Scene Understanding for Autonomous Mobile Robots in Warehouse Operations
14:20-14:40 , Microblink
Transformer Architectures in Production: Efficiency and Self-Supervision
14:40-15:00 , Xylon
Implementing Computer Vision Algorithms on FPGAs: Challenges & Advantages
15:00-15:20 , RoMB technologies
From the Road into the Warehouse: Semantic Segmentation for Autonomous Forklifts
15:20-15:35 Coffee Break
15:35-15:50 , University of Zagreb
Cross-Domain Learning of Dense Prediction Models
15:50-16:05 , University of Zagreb
Dense Semantic Forecasting
16:05-16:20 , University of Zagreb
Dense Anomaly Detection with Synthetic Negatives
16:20-16:35 , University of Zagreb
Revisiting One-Way Consistency for Semi-Supervised Semantic Segmentation
16:35-16:50 , University of Zagreb
Detection of Road-Safety Attributes in Video


The workshop has been supported by the University of Wuppertal and the Croatian Science Foundation project ADEPT.