CycleCrash: A Dataset of Bicycle Collision Videos for Collision Prediction and Analysis

TL;DR

CycleCrash is a comprehensive dataset of 3,000 cyclist collision videos designed to advance collision prediction and analysis, complemented by a novel VidNeXt model leveraging ConvNeXt and transformers for improved video understanding.

Contribution

The paper introduces CycleCrash, a new dataset for cyclist collision analysis, and proposes VidNeXt, a novel model combining ConvNeXt and non-stationary transformers for video-based collision prediction.

Findings

VidNeXt outperforms 6 baseline models on CycleCrash tasks.

The dataset enables 9 different cyclist collision prediction tasks.

Detailed ablation studies validate the effectiveness of VidNeXt.

Abstract

Self-driving research often underrepresents cyclist collisions and safety. To address this, we present CycleCrash, a novel dataset consisting of 3,000 dashcam videos with 436,347 frames that capture cyclists in a range of critical situations, from collisions to safe interactions. This dataset enables 9 different cyclist collision prediction and classification tasks focusing on potentially hazardous conditions for cyclists and is annotated with collision-related, cyclist-related, and scene-related labels. Next, we propose VidNeXt, a novel method that leverages a ConvNeXt spatial encoder and a non-stationary transformer to capture the temporal dynamics of videos for the tasks defined in our dataset. To demonstrate the effectiveness of our method and create additional baselines on CycleCrash, we apply and compare 7 models along with a detailed ablation. We release the dataset and code at https://github.com/DeSinister/CycleCrash/ .