BADAS: Context Aware Collision Prediction Using Real-World Dashcam Data

2025-10-16

AI Summary

Existing forward collision warning (FCW) systems struggle to distinguish genuine threats involving the ego-vehicle from irrelevant surrounding incidents, resulting in high false-positive rates. To address this, we propose BADAS, an ego-vehicle-centric collision prediction framework, and introduce the first benchmark dedicated to ego-vehicle collision assessment—comprising a re-annotated dataset, consensus-based warning-time labels, and open-source models. Methodologically, BADAS employs an end-to-end trained V-JEPA2 backbone, trained on real-world driving videos to yield two variants: BADAS-Open and BADAS 1.0. Evaluated across multiple mainstream benchmarks, BADAS substantially outperforms conventional ADAS baselines, achieving state-of-the-art performance in average precision (AP) and area under the ROC curve (AUC). Moreover, it delivers significantly more accurate collision countdown predictions, effectively reducing false alarms while maintaining high sensitivity to imminent ego-vehicle collisions.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Multiagent Systems: Adversarial Agents
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Existing collision prediction methods often fail to distinguish between ego-vehicle threats and random accidents not involving the ego vehicle, leading to excessive false alerts in real-world deployment. We present BADAS, a family of collision prediction models trained on Nexar's real-world dashcam collision dataset -- the first benchmark designed explicitly for ego-centric evaluation. We re-annotate major benchmarks to identify ego involvement, add consensus alert-time labels, and synthesize negatives where needed, enabling fair AP/AUC and temporal evaluation. BADAS uses a V-JEPA2 backbone trained end-to-end and comes in two variants: BADAS-Open (trained on our 1.5k public videos) and BADAS1.0 (trained on 40k proprietary videos). Across DAD, DADA-2000, DoTA, and Nexar, BADAS achieves state-of-the-art AP/AUC and outperforms a forward-collision ADAS baseline while producing more realistic time-to-accident estimates. We release our BADAS-Open model weights and code, along with re-annotations of all evaluation datasets to promote ego-centric collision prediction research.

Problem

• Distinguishes ego-vehicle threats from irrelevant accidents
• Reduces false alerts in real-world collision prediction systems
• Enables fair evaluation using re-annotated ego-centric benchmarks

Innovation

• Uses V-JEPA2 backbone for end-to-end training
• Re-annotates benchmarks for ego-involvement identification
• Synthesizes negatives and adds consensus alert-time labels