Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CADD: A Chinese Traffic Accident Dataset for Statute-Based Liability Attribution

Yunfei Shen, Zhongcheng Wu

TL;DR

This work tackles the need for legally grounded liability attribution in autonomous driving by introducing the Chinese Accident Duty-determination Dataset (CADD), the first benchmark that links observable driving behaviors to liability judgments and specific Chinese traffic-law statutes via a Behavior–Liability–Statute pipeline. CADD comprises 792 real-world dashcam videos with hierarchical, symmetric annotations, definitive liability judgments, and explicit statute mappings, enabling evaluation of automated liability attribution, causal reconstruction, and explainable legal reasoning. The authors provide a comprehensive data collection and curation workflow, an annotation framework, quality-control procedures, and rich data analyses to demonstrate the dataset’s potential for perception-to-legal reasoning. The work paves the way for legally compliant and accountable autonomous systems by enabling temporally grounded, explainable, and statute-aware decision-making in traffic incidents.

Abstract

As autonomous driving technology advances, the critical challenge evolves beyond collision avoidance to the \textbf{adjudication of liability} when accidents occur. Existing datasets, focused on detection and localization, lack the annotations required for this legal reasoning. To bridge this gap, we introduce the \textbf{C}hinese \textbf{A}ccident \textbf{D}uty-determination \textbf{D}ataset (\textbf{CADD}), the first benchmark for statute-based liability attribution. CADD contains 792 real-world driving recorder videos, each annotated within a novel \textbf{``Behavior--Liability--Statute''} pipeline. This framework provides \textbf{granular, symmetric behavior annotations}, clear responsibility assignments, and, uniquely, links each case to the specific \textbf{Chinese traffic law statute} violated. We demonstrate the utility of CADD through detailed analysis and establish benchmarks for liability prediction and explainable decision-making. By directly connecting perceptual data to legal consequences, CADD provides a foundational resource for developing accountable and legally-grounded autonomous systems.

CADD: A Chinese Traffic Accident Dataset for Statute-Based Liability Attribution

TL;DR

This work tackles the need for legally grounded liability attribution in autonomous driving by introducing the Chinese Accident Duty-determination Dataset (CADD), the first benchmark that links observable driving behaviors to liability judgments and specific Chinese traffic-law statutes via a Behavior–Liability–Statute pipeline. CADD comprises 792 real-world dashcam videos with hierarchical, symmetric annotations, definitive liability judgments, and explicit statute mappings, enabling evaluation of automated liability attribution, causal reconstruction, and explainable legal reasoning. The authors provide a comprehensive data collection and curation workflow, an annotation framework, quality-control procedures, and rich data analyses to demonstrate the dataset’s potential for perception-to-legal reasoning. The work paves the way for legally compliant and accountable autonomous systems by enabling temporally grounded, explainable, and statute-aware decision-making in traffic incidents.

Abstract

As autonomous driving technology advances, the critical challenge evolves beyond collision avoidance to the \textbf{adjudication of liability} when accidents occur. Existing datasets, focused on detection and localization, lack the annotations required for this legal reasoning. To bridge this gap, we introduce the \textbf{C}hinese \textbf{A}ccident \textbf{D}uty-determination \textbf{D}ataset (\textbf{CADD}), the first benchmark for statute-based liability attribution. CADD contains 792 real-world driving recorder videos, each annotated within a novel \textbf{``Behavior--Liability--Statute''} pipeline. This framework provides \textbf{granular, symmetric behavior annotations}, clear responsibility assignments, and, uniquely, links each case to the specific \textbf{Chinese traffic law statute} violated. We demonstrate the utility of CADD through detailed analysis and establish benchmarks for liability prediction and explainable decision-making. By directly connecting perceptual data to legal consequences, CADD provides a foundational resource for developing accountable and legally-grounded autonomous systems.

Paper Structure

This paper contains 18 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Traffic Accident Datasets
  4. Driving Behavior and Risk Assessment
  5. Legal Reasoning and Autonomous Vehicle Liability
  6. The Chinese Accident Duty-determination (CADD) Dataset
  7. Data Collection and Curation
  8. Annotation Framework
  9. Annotation Process and Quality Control
  10. Dataset Statistics and Analysis
  11. Data Overview and Case Studies
  12. Potential Applications and Benchmarking
  13. Automated Liability Attribution
  14. Legal Statute Identification
  15. Temporal and Causal Analysis
  16. ...and 3 more sections

Figures (3)

  • Figure 1: Detailed annotation statistics of the CADD dataset across nine dimensions: (i) scene types, (ii) road types, (iii) accident types, (iv) ego vehicle behavior, (v) other vehicle behavior, (vi) traffic violation statutes, (vii) collision types, (viii) liability attribution, and (ix) view types.
  • Figure 2: Correlation analysis of traffic accident factors in the CADD dataset: (i) collision types × road types, (ii) collision types × scene types, (iii) accident types × road types, (iv) collision types × accident types, (v) collision types × legal statutes, and (vi) accident types × legal statutes.
  • Figure 3: Annotated accident examples with paired image–text presentation. Each example includes a composite image (top) and compact two-column annotation summary (bottom). (a–f) represent diverse accident types across different environments, behaviors, and legal outcomes.