Dahua Launches Xinghan AI Models to Drive Next-Generation Intelligent AIoT Solutions

Dahua Technology, a global leader in video-centric AIoT solutions and services, has introduced its Xinghan Large-scale AI Models, an advanced, industry-grade AI system combining large-scale visual intelligence with multimodal and language capabilities. Designed to tackle complex real-world challenges, Xinghan marks a significant step forward in Dahua’s innovation journey, enabling intelligent transformation across multiple sectors.

Technological Foundation of Xinghan

Inspired by the Chinese word for “galaxy,” Xinghan bridges cutting-edge research with practical applications, offering full-stack capabilities powered by edge-cloud synergy for scalable, adaptive intelligence. The Xinghan architecture includes three core model series: L, V, and M. The L-series focuses on natural language understanding and interaction, while the V- and M-series target specialized applications.

V-Series: Xinghan Vision Models

Designed for advanced visual intelligence and video analytics, the V-Series simplifies model complexity by concentrating on key targets such as humans, motor vehicles, and non-motor vehicles, while maintaining high accuracy.

  • Perimeter Protection: Extends coverage by detecting smaller targets (as small as 20×20 pixels), reducing false alarms and increasing the detection range of large-model cameras.*
  • WizTracking: Next-generation intelligent tracking algorithm handling complex occlusions and variations in target posture, improving accuracy by 50%.*
  • Crowd Map: Enhances small-target detection at long distances (up to 2× farther), features umbrella compensation, improves accuracy by 80% in rainy conditions*, supports detection of up to 5,000 people, and performs robustly in dense crowds and low-light environments.*
  • Scene Adaptive – AI WDR: Analyzes spatial and contextual characteristics to enable intelligent, automated camera configuration.
  • AI Rule Assist: Automatically delineates Perimeter Protection intrusion rules with one-click setup, accurate scene recognition, and automatic analysis.

M-Series: Xinghan Multimodal Models

Multimodal models process and integrate multiple heterogeneous data types such as text, images, audio, and video. This enhances information processing efficiency, enables more natural human-computer interaction, and unlocks diverse application scenarios.

  • WizSeek: Revolutionizes video investigation via natural language search. Describe your target (e.g., people, vehicle, animal, or item), and WizSeek retrieves matching footage from video archives instantly.
  • Text-Defined Alarms: Allows users to define alarms using natural language, lowering development thresholds and enabling fast, flexible, and scalable configuration for varied real-world scenarios.

Explore Itech360hub.com for the latest updates on AI, IoT, cybersecurity, and insights from industry experts.