SDAD 2026 @ ECCV

Safe and Defensive Autonomous Driving Workshop on ECCV 2026

Building autonomous systems that proactively avoid latent hazards, reason under uncertainty, and deliver trustworthy defensive behavior.

Submit to OpenReview Accepted Papers Call for Papers View Program

About

SDAD (Safe and Defensive Autonomous Driving) is an ECCV 2026 full-day workshop in Malmö. We organize around three core problems:

  1. What dangers appear in driving scenes—including latent hazards that are difficult to observe?
  2. How can autonomous vehicles learn defensive driving under uncertainty and long-tail situations?
  3. How can we evaluate defensive driving behaviors beyond trajectory-centric benchmarks?

Program

6 keynote talks, an oral session with 3 papers, poster sessions, a panel discussion, and the best paper awards.

Morning

  • 09:00–09:15 Welcome Speech
  • 09:15–09:45 Keynote Talk 1 Andreas Geiger
  • 09:45–10:15 Keynote Talk 2 Dragomir Anguelov
  • 10:15–10:30 Coffee break
  • 10:30–11:15 Oral Session (3 × 15 min)
  • 11:15–12:15 Poster Session
  • 12:15–14:00 Lunch Break and Poster Session

Afternoon

  • 14:00–14:30 Keynote Talk 3 Chong Ruan
  • 14:30–15:00 Keynote Talk 4 Dan Levi
  • 15:00–15:30 Coffee break
  • 15:30–16:00 Keynote Talk 5 Boyi Li
  • 16:00–16:30 Keynote Talk 6 Jun Gao
  • 16:30–17:15 Panel Discussion Host: Qifeng Chen
  • 17:15–17:30 Award and Closing Remark

Keynote Speakers

Portrait of Andreas Geiger

Andreas Geiger

Professor of Computer Science

University of Tübingen

Portrait of Dragomir Anguelov

Dragomir Anguelov

VP and Head of Research

Waymo

Portrait of Jun Gao

Jun Gao

Assistant Professor

University of Michigan

Portrait of Chong Ruan

Chong Ruan

Chief Scientist

DeepRoute.Ai

Portrait of Boyi Li

Boyi Li

Postdoctoral Researcher

UC Berkeley

Research Scientist

NVIDIA Research

Portrait of Dan Levi

Dan Levi

Foundational AI Research Group Lead

General Motors R&D

Call for Papers

We invite non-archival submissions on defensive driving, latent hazards, learning under uncertainty, and evaluation beyond trajectory-centric benchmarks. Reviews run on OpenReview; accepted work is presented as posters, with oral and best-paper selections.

Submission deadline: July 15, 2026

Submit to OpenReview Dates, rules & review

Accepted Papers

Eighteen papers were accepted to SDAD 2026, three of them selected for oral presentations. Together they span latent hazard anticipation, planning under occlusion and uncertainty, critical scenario generation, and evaluation protocols that go beyond trajectory-centric benchmarks.

Abstracts and camera-ready PDFs for each accepted paper.

Browse accepted papers

Organizers

Portrait of Qifeng Chen

Qifeng Chen

Associate Professor

HKUST

Portrait of Alexandre Alahi

Alexandre Alahi

Associate Professor

EPFL

Portrait of Bo Li

Bo Li

Associate Professor

UIUC

Portrait of Yiyi Liao

Yiyi Liao

Assistant Professor

Zhejiang University

Portrait of Li Erran Li

Li Erran Li

Head of Science

AWS AI

Portrait of Tongyi Cao

Tongyi Cao

CTO

DeepRoute.Ai

Student Co-Organizers

  • Zhili Chen PhD Student HKUST
  • Qiang Wen Researcher DeepRoute.Ai
  • Jiaqi Tang PhD Student HKUST
  • I Chieh Chen PhD Student HKUST

Sponsors

The workshop is gratefully sponsored by DeepRoute (“元戎启行”).