awesome_explainable_driving

A curated list of papers on explainability and interpretability of self-driving models

Most of the references below are organized and discuss in the following survey:

• Explainability of vision-based autonomous driving systems: Review and challenges (submitted to IJCV), Eloi Zablocki, Hédi Ben-Younes, Patrick Pérez, Matthieu Cord [arxiv]

Table of Contents

• Saliency maps
• Counterfactual interventions and causal inference
• Representation
• Evaluation
• Attention maps
• Semantic inputs
• Predicting intermediate representations
• Output interpretability
• Natural language explanations
• Datasets

Saliency maps

• Explaining How a Deep Neural Network Trained with End-to-End Learning Steers a Car (2017, arxiv), Mariusz Bojarski, Philip Yeres, Anna Choromanska, Krzysztof Choromanski, Bernhard Firner, Lawrence Jackel, Urs Muller [arxiv]
• Visualbackprop: Efficient visualization of cnns for au-tonomous driving (2018, ICRA), Mariusz Bojarski, Anna Choromanska, Krzysztof Choromanski, Bernhard Firner, Larry Jackel, Urs Muller, Karol Zieba [arxiv]
• Interpretable learning for self-driving cars by visualizing causal attention (2017, ICCV), Jinkyu Kim, John Canny [arxiv]
• Conditional Affordance Learning for Driving in Urban Environments (2018, CoRL), Axel Sauer, Nikolay Savinov, Andreas Geiger [arxiv]
• Interpretable Self-Attention Temporal Reasoning for Driving Behavior Understanding (2020, ICASSP), Yi-Chieh Liu, Yung-An Hsieh, Min-Hung Chen, Chao-Han Huck Yang, Jesper Tegner, Yi-Chang James Tsai [arxiv]

Counterfactual interventions and causal inference

• Explaining How a Deep Neural Network Trained with End-to-End Learning Steers a Car (2017, arxiv), Mariusz Bojarski, Philip Yeres, Anna Choromanska, Krzysztof Choromanski, Bernhard Firner, Lawrence Jackel, Urs Muller [arxiv]
• Who Make Drivers Stop? Towards Driver-centric Risk Assessment: Risk Object Identification via Causal Inference (2020, IROS), Chengxi Li, Stanley H. Chan, Yi-Ting Chen [arxiv]
• ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst (2019 Robotics, Science and Systems), Mayank Bansal, Alex Krizhevsky, Abhijit Ogale [arxiv]

Representation

• DeepTest: Automated Testing of Deep-Neural-Network-driven Autonomous Cars (2018, ICSE), Yuchi Tian, Kexin Pei, Suman Jana, Baishakhi Ray [arxiv]

Evaluation

• ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst (2019 Robotics, Science and Systems), Mayank Bansal, Alex Krizhevsky, Abhijit Ogale [arxiv]
• Learning Accurate and Human-Like Driving using Semantic Maps and Attention (2020, IROS), Simon Hecker, Dengxin Dai, Alexander Liniger, Luc Van Gool [arxiv]
• DeepTest: Automated Testing of Deep-Neural-Network-driven Autonomous Cars (2018, ICSE), Yuchi Tian, Kexin Pei, Suman Jana, Baishakhi Ray [arxiv]

Attention maps

• Interpretable learning for self-driving cars by visualizing causal attention (2017, ICCV), Jinkyu Kim, John Canny [arxiv]
• Deep Object-Centric Policies for Autonomous Driving (2019, ICRA), Dequan Wang, Coline Devin, Qi-Zhi Cai, Fisher Yu, Trevor Darrell [arxiv]
• Attentional Bottleneck: Towards an Interpretable Deep Driving Network (2020, arxiv), Jinkyu Kim, Mayank Bansal [arxiv]
• Learning Accurate and Human-Like Driving using Semantic Maps and Attention (2020, IROS), Simon Hecker, Dengxin Dai, Alexander Liniger, Luc Van Gool [arxiv]

Semantic inputs

Predicting intermediate representations

Output interpretability

Natural language explanations

Datasets