SB-SIF: Scenario Based - Synthetic Intersection Footage

Repository of Scenario Based - Synthetic Intersection Footage (SB-SIF) used to develop an automatic accident detection system in the paper "YoFlow Method for Scenario Based Automatic Accident Detection" in 2025.     The dataset presents synthetically generated traffic footage created using the PEGASUS scenario-based approach. It is based on five common intersection crash scenarios derived from German accident data reported in the literature (section 3 of research paper). These five scenarios were parameterized using BeamNG.tech v0.30 simulation platform to generate 914 crash videos, 123 near-miss events, and 924 instances of normal traffic footage.    Annotation note:  Footage instances are annotated on video level using three formats. Non-crash instances: (NT-XX),  Near-miss instances: (NM-XX). Crash instances: (XX-XX) , which indicates EGO vehicle speed - Crashed Vehicle speed runned during the simulation.