Dahua Technology, the world's leading video-centric AIOT solutions and service provider, has officially launched the large-scale AI model of Xinghan, the next-generation industry-grade AI system that integrates large-scale visual intelligence with multimodal and language capabilities. Developed to address the complex challenges of the real world environment, Xinghan represents a major leap in Duffa's ongoing innovation, enhancing intelligent transformation across diverse sectors.
Also Read: Aithority Interview with Dr. Petar Tsankov, CEO and co-founder of LatticeFlow AI
Xinghan's Technical Foundation
With its mission to enable machines to truly understand the world, Xinghan model systems continue to evolve by bridging cutting-edge research in real-world applications. Named after the Chinese word “Galaxy,” Xinghan offers a full stack feature matrix powered by Edge-Cloud Synergy, enabling scalable adaptive intelligence across the industry. The upgraded Xinghan architecture consists of three core model series. L, V, M.
V Series: Xinghan Vision Models
Focusing on advanced visual intelligence and video analytics, the series streamlines target categories by focusing on key targets (human, car, non-moving vehicles, etc.) to reduce model complexity while maintaining high accuracy.
The main features are:
- Boundary Protection: Coverage is extended by accurately identifying smaller targets (even up to 20 x 20 pixels) compared to traditional CNN-based AI models, reducing false alarms and increasing detection range for large cameras.
- With Tracking: It provides next-generation intelligent tracking algorithms, dealing with complex blockages and target pose variations, and improves accuracy by 50%. *
- Cloud Map: It greatly promotes small target detection over long distances (up to 2x far), features umbrella compensation, improving accuracy by 80% in rainy conditions*. It also increases the analysis range by 2.5x, supporting detection of up to 5,000 people, providing robust performance in dense crowds and low light environments.
- Scene Adaptive – AI WDR: It leverages situational awareness to analyze both the spatial and contextual characteristics of the scene, enabling intelligent and automated camera configurations.
- AI Rule Assist: It is designed for automatic depiction of boundary protection intrusion rules, providing one-click access, highly accurate scene recognition, automatic analysis and more.
Also Read: Developing Autonomous Security Agents Using Computer Vision and Generation AI
M Series: Xinghan Multimodal Model
Multimodal models are advanced AI systems that can simultaneously process and deeply integrate multiple heterogeneous data types (text, images, audio, video, etc.). This greatly improves the efficiency of information processing, allows for more natural human computer interactions, and unlocks a wider range of application scenarios.
The main features are:
- wizseek: Revolutionize video research through natural language search. Simply explain your target (people, vehicle, animal, item, etc.), and Wizseek instantly gets matching footage across the recorded video archive.
- TextDefined alarms: This allows users to define alarms simply by describing them in natural language. This significantly lowers development thresholds and allows for fast, flexible, scalable configurations tailored to a variety of real-world scenarios.
[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]
