In-distribution

A standardized benchmark based on a stratified split of the real dataset.

Benchmark overview

The in-distribution benchmark acts as the standardized ACCIDENT setting. It is built from a stratified split of the real dataset so that key attributes, including collision type and major scene characteristics, remain as similar as possible across training and test, minimizing unintended dataset shift.

Tasks: temporal localization, spatial localization, and collision type classification.
Protocol: controlled evaluation on a 507 / 1,520 train-test split of the real dataset.
Official score: unified ACCIDENT score together with the three task-specific metrics.

Score Your Submission

Select CSV File

Leaderboard

Ranked by unified score

The public table reports per-task scores and links to the submitting method or paper when available. Click a metric header to sort the ranking.

	Method	Paper

No scored submissions yet.