02/24 2025
442
A new era in multi-modal large models has dawned with the latest introduction by Hikvision. The company has launched a series of text search computing products, leveraging the Guanlan large model technology system, spanning from edge to center. This includes the Text Search Super Brain and Text Search Host. By inputting a sentence or a word, users can achieve efficient search and real-time alerts, offering enhanced recognition, more precise search capabilities, and flexible applications. This aids in boosting intelligence and efficiency in security business management across diverse industries.
Drawing on its extensive experience in the security field, Hikvision integrates its proprietary multi-modal large models with intelligent computing hardware platforms. This results in comprehensive upgrades across intelligent computing, storage, search, and application functions, enabling multi-modal fusion computing and cross-modal search applications. With the support of large models, intelligent computing products have evolved from offering specific functions to providing open capabilities, thereby more effectively meeting users' diverse management needs.
Enhanced Recognition: Large Model Detection of All Entities with Heightened Perception
The intelligent computing product series is equipped with a large model capable of detecting a wide range of entities, significantly boosting detection and recognition abilities. Whether it's moving individuals, vehicles, animals, or stationary objects like billboards, trash cans, flowers, trees, and grass in video footage, they can all be recognized, achieving a more holistic perception.
More Precise Search: Deep Semantic Understanding and High-Precision Secondary Search
The products feature multi-modal large models with deep semantic understanding, extracting features from images and search languages. By aligning image features with key semantic features in the search language, cross-modal high-precision search is enabled. For instance, it can comprehend text descriptions of targets like "orange cat," "overflowing trash can," "excessive stacking on a vehicle," and "working at heights," and accurately search for related targets. Users can input a sentence or a word to search for target images within seconds, not limited to people and vehicles but extending to "everything" and events, exponentially enhancing target search efficiency. (Data sourced from Hikvision Labs)
Furthermore, the product supports online fine-tuning of search results, allowing users to train the model on-site, improving accuracy with each use.
Flexible Application: Custom Text Alerts to Meet Diverse Needs
Leveraging the large model's ability to fuse and process multiple modal information, the text search computing product series can provide custom text alerts. By inputting a sentence or a word, real-time alerts and reminders can be generated for specific events and targets, such as "two people riding an electric bicycle," "overflowing trash can," and "crossing the guardrail," making management more prompt and effective.
Additionally, system deployment is straightforward and swift. Users can establish an intelligent system by utilizing the text search computing product series alongside basic security software and hardware combinations. Whether for upgrades, reconstructions, or building new security systems, flexible deployment is achievable, reducing general deployment time from days to hours.
The evolution of Hikvision's Guanlan large model technology system drives the continuous implementation of intelligent security products. Benefiting from the robust perception and cross-task generalization capabilities of the Guanlan visual large model and the image-text multi-modal large model, the intelligentization of more security scenarios will be further accelerated.
Through model structure design, knowledge distillation, model simplification, and other large model deployment technologies tailored for various platforms, along with innovative designs for intelligent hardware, Hikvision has made large models accessible on diverse computing power hardware platforms. This allows large model technology to genuinely meet the security application needs of various industries and scenarios.
Relying on Hikvision's Guanlan large model technology system, the text search storage and text search computing product series are propelling security into a smarter and more efficient era.