DADA-2000: Can Driving Accident be Predicted by Driver Attentionƒ Analyzed by A Benchmark

Abstract

Driver attention prediction is currently becoming the focus in safe driving research community, such as the DR(eye)VE project and newly emerged Berkeley DeepDrive Attention (BDD-A) database in critical situations. In safe driving, an essential task is to predict the incoming accidents as early as possible. BDD-A was aware of this problem and collected the driver attention in laboratory because of the rarity of such scenes. Nevertheless, BDD-A focuses the critical situations which do not encounter actual accidents, and just faces the driver attention prediction task, without a close step for accident prediction. In contrast to this, we explore the view of drivers' eyes for capturing multiple kinds of accidents, and construct a more diverse and larger video benchmark than ever before with the driver attention and the driving accident annotation simultaneously (named as DADA-2000), which has 2000 video clips owning about 658, 476 frames on 54 kinds of accidents. These clips are crowd-sourced and captured in various occasions (highway, urban, rural, and tunnel), weather (sunny, rainy and snowy) and light conditions (daytime and nighttime). For the driver attention representation, we collect the maps of fixations, saccade scan path and focusing time. The accidents are annotated by their categories, the accident window in clips and spatial locations of the crash-objects. Based on the analysis, we obtain a quantitative and positive answer for the question in this paper.

Driving Video Fixation Prediction Model Via Spatio-Temporal Networks and Attention Gates

“…Berkley DeepDrive laboratory published a large-scale in-lab driver attention dataset, Berkeley Deep-Drive Attention (BDD-A) dataset, which is built upon braking event videos selected from a large-scale, crowd-sourced driving video dataset [11], and proposed a model to predict the driver attention. Besides, Fang et al [12] constructed a new benchmark with 2000 video sequences (named as DADA-2000) with 658,476 frames, laboriously annotated the driver attention and accident intervals. Deng et al built a static image dataset including 40 drivers' eye-tracking data [13], and then proposed an saliency detection model based on random forest [14].…”

Uncertainty-based Traffic Accident Anticipation with Spatio-Temporal Relational Learning

“…Accident anticipation aims to predict an accident from dashcam video before it happens. It is one of the most important tasks for safety-guaranteed autonomous driving applications and has been receiving increasing attentions in recent years [4,7,10,32]. Thanks to accident anticipation, the safety level of intelligent systems on vehicles could be significantly enhanced.…”

Gaze Focalization System for Driving Applications Using OpenFace 2.0 Toolkit with NARMAX Algorithm in Accidental Scenarios

“…Usually, visual attention systems are validated in simulation with simple scenes. However, the simulation of driver attention in complex driving scenarios is rather challenging and highly subjective [ 3 ]. These systems aim to automatically estimate the area where the driver is focusing their attention.…”

“…However, these scenes are not challenging enough to test the attention because these situations do not cause accidents. Therefore, we have evaluated our system on the recent DADA2000 database [ 3 ], which is the largest and most diverse driving attention database composed by accidental scenarios. It contains 2000 video clips annotated with the accident localization and the attention map generated in-lab by some users with fairly complex accidental scenarios in diverse weather and lighting conditions.…”