#China #Changzhou #Dahua Technology #AI Models #Xinghan AI

Continuous Evolution for an Intelligent Future of AIoT

Dahua Technology, a leading global provider of AIoT solutions focused on video, officially introduced its extensive Xinghan AI models. These next-generation AI systems integrate advanced visual intelligence with multimodal and language capabilities, tailored to meet various industry demands. The Xinghan system represents a significant leap forward for Dahua in addressing real-world complexities and facilitating intelligent transformations across different sectors.

Foundation of the Xinghan Technology

The Dahua Xinghan Large-scale AI Models are built around visual analysis and leverage multimodal capabilities. By incorporating deep industry expertise, they create a customized model for varied application scenarios. This innovative approach is grounded in real-world requirements, acting as the key to realizing scalable and commercially viable AI solutions.

The goal of the Xinghan system is to enable machines to genuinely understand their surroundings. Hence, Dahua is continually evolving this model, amalgamating cutting-edge research with practical applications. The name 'Xinghan,' which translates to 'galaxy' in Chinese, reflects a comprehensive skillset based on edge-cloud synergy, providing scalable and adaptive intelligence across industries.

Key Features of the Xinghan Models

The modernized architecture of Xinghan encompasses three primary model series: L, V, and M.

• - Series L focuses on natural language understanding and interaction.
• - Series V, named the Xinghan Vision Models, specializes in advanced visual intelligence and video analysis. This series simplifies target categorization by concentrating on key elements such as people, vehicles, and non-motorized objects to reduce the complexity of the model while retaining high accuracy.
• - Features of the Vision Models include:

- Perimeter Protection: Enhances coverage by accurately identifying smaller targets (as low as 20 × 20 pixels), reducing false alarms while increasing camera detection range with the large model.
- WizTracking: Employs an intelligent tracking algorithm capable of handling complex occlusions and target position changes, achieving a 50% improvement in precision.
- Crowd Map: Significantly enhances small target detection from longer distances (up to twice farther), equipped with rain compensation features that boost accuracy by 80% during adverse weather. It supports the detection of up to 5,000 individuals, maintaining robust performance in crowded environments and low-light conditions.
- Scene Adaptive – AI WDR: Utilizes situational awareness to analyze spatial and contextual scene characteristics, facilitating intelligent and automated camera configuration.
- AI Rule Assistant: Designed for automatic perimeter protection rule designation, offering one-click access, highly precise scene recognition, and automated analysis.

• - Series M consists of multimodal models capable of processing and deeply integrating multiple heterogeneous data types (e.g., text, images, audio, and video) simultaneously. This advancement significantly enhances information processing efficiency, enabling more natural human-computer interaction and broadening the uses of the system.
• - Noteworthy features include:

- WizSeek: Revolutionizes video searching via natural language requests. Users can simply describe a target (such as a person, vehicle, animal, or object), and the WizSeek application will promptly retrieve corresponding footage from archived videos.
- Text-defined Alarms: Allows users to set alarm parameters through simple descriptive phrases in natural language, greatly reducing development demands while facilitating quick, flexible, and scalable configurations tailored to various real-world scenarios.

For more details on the large-scale AI models offered by Dahua, check here for images and here for the company logo. The results mentioned herein are based on standard settings and testing environments.