GTA-Crime: A Synthetic Dataset and Generation Framework for Fatal Violence Detection with Adversarial Snippet-Level Domain Adaptation

2025-09-09

AI Summary

Detecting lethal violent events (e.g., shootings, stabbings) in surveillance videos remains challenging due to the scarcity of real-world annotated data and the high cost of labeling. To address this, we propose GTA-Crime—the first synthetic, multi-view, multi-scenario dataset specifically designed for lethal violence recognition, generated using the *Grand Theft Auto V* game engine. To mitigate domain shift between synthetic and real surveillance footage, we introduce a clip-level Wasserstein adversarial domain adaptation method that enables fine-grained feature alignment across domains. Extensive experiments demonstrate that our synthetic dataset, coupled with the proposed domain adaptation strategy, significantly improves detection accuracy on real-world benchmarks such as UCF-Crime. These results validate both the efficacy of our synthetic data generation framework and the generalizability of our adversarial transfer learning approach for violence detection.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Data transparency and provenance

Abstract

Recent advancements in video anomaly detection (VAD) have enabled identification of various criminal activities in surveillance videos, but detecting fatal incidents such as shootings and stabbings remains difficult due to their rarity and ethical issues in data collection. Recognizing this limitation, we introduce GTA-Crime, a fatal video anomaly dataset and generation framework using Grand Theft Auto 5 (GTA5). Our dataset contains fatal situations such as shootings and stabbings, captured from CCTV multiview perspectives under diverse conditions including action types, weather, time of day, and viewpoints. To address the rarity of such scenarios, we also release a framework for generating these types of videos. Additionally, we propose a snippet-level domain adaptation strategy using Wasserstein adversarial training to bridge the gap between synthetic GTA-Crime features and real-world features like UCF-Crime. Experimental results validate our GTA-Crime dataset and demonstrate that incorporating GTA-Crime with our domain adaptation strategy consistently enhances real world fatal violence detection accuracy. Our dataset and the data generation framework are publicly available at https://github.com/ta-ho/GTA-Crime.

Problem

• Detecting rare fatal violence incidents in videos
• Bridging synthetic-real domain gap for violence detection
• Generating synthetic fatal violence video dataset

Innovation

• Synthetic dataset from GTA5 game
• Snippet-level domain adaptation strategy
• Wasserstein adversarial training for feature alignment