2026/11/1320
International Data Corporation (IDC) recently released its China AI Software Market Semiannual Tracker, 1H26. The report shows that China’s on-premises computer vision market grew 16.8% year-on-year to RMB4.32 billion in the first half of 2026. SenseTime ranked first with a 19.5% market share. Supported by its deep expertise in vision algorithms and city-scale visual AI platforms, the Company maintained its leadership in public safety, urban governance, and other applications.
SenseTime ranked first by market share in IDC’s reports on China’s visual AI market for 10 consecutive years from 2016 to 2025. Its latest top ranking in the new assessment cycle underscores its sustained competitiveness in visual AI.
As AI technology continues to advance rapidly, visual AI is entering a new phase. IDC notes that large models do more than replace traditional AI capabilities. By enhancing complex scene understanding, business interaction, and task orchestration, they complement specialized models and broaden the scope of AI applications. With CV 2.0 as a core strategic focus, SenseTime is integrating visual AI with multimodal large models and AI agents to advance its Agentic CV approach.
From Seeing to Understanding and Acting: Visual AI Enters CV 2.0
SenseTime has built deep expertise across vision algorithms, model development, product platforms, and industry applications through its sustained focus on computer vision. It is now advancing visual AI through CV 2.0. While traditional visual AI primarily detects and recognizes objects and events in images and video, CV 2.0 enables complex scene understanding, cross-modal analysis, and coordinated task execution. This moves visual intelligence beyond individual algorithm applications toward end-to-end intelligent workflows built around business objectives.
SenseTime’s SenseFoundry has established a comprehensive CV 2.0 visual intelligence architecture. Specialized vision models perform tasks such as lightweight object recognition and efficient object retrieval, while the one-stop Monolith algorithm production platform enables the rapid training, deployment, and management of vision models and algorithms. Powered by SenseNova, SenseTime’s family of multimodal large models, the architecture integrates visual, textual, and business data to enhance complex scene understanding and reasoning. Multimodal agent infrastructure supports vision algorithm training, tool use, and coordinated agent execution, extending visual AI from perception and analysis to task execution.
Building on these capabilities, SenseFoundry offers a product portfolio spanning urban governance, public safety, commercial operations, and government administration. The portfolio includes city and enterprise AI enablement platforms, the Nebula edge product series, and SenseTime’s Raccoon Government Data Edition (小浣熊·政数版), embedding visual AI capabilities into city operations and enterprise management.
Putting Vast Visual Datasets to Work: From Passive Recognition to Proactive Action
Unlocking greater value from visual AI depends on the ability to quickly extract useful information from the vast image and video datasets generated by cities and enterprises, understand complex events, and initiate follow-up actions.
Urban flood prevention illustrates this shift, enabling a more proactive workflow spanning risk detection, early warning, and response.
During typhoons and other extreme weather events, city emergency centers need to use weather forecasts to identify potential hazards in advance, including flood-prone locations, fallen trees, and falling objects, and promptly alert the responsible agencies. Previously, a single operator might need to monitor dozens of video feeds. When an incident occurred, operators had to review recordings from hundreds of feeds individually, creating a heavy workload and limiting response efficiency.
SenseFoundry’s city AI enablement platform transforms risk monitoring from manual inspection into proactive system-level understanding and intelligent coordination. Through citywide spatiotemporal modeling, the system connects information on people, vehicles, events, objects, and locations to map risk hotspots and analyze patterns in recurring disasters. Intelligent video tagging and natural-language search turn existing footage into searchable, analyzable data assets, enabling users to quickly locate relevant sites and clips. Automated detection, filtering, and alerts reduce prolonged screen monitoring and repetitive checks, allowing operators to focus on confirming alerts and coordinating responses. The platform also enables video resource sharing and operational coordination across departments, allowing monitoring resources managed by different departments to support multiple use cases.
Building on this foundation, SenseTime’s Raccoon Government Data Edition, designed for government and enterprise users, enables more flexible and efficient execution of flood-prevention tasks. Staff can express their requirements in natural language, while agents interpret task objectives and dynamically orchestrate and invoke tools for information retrieval, data analysis, and report generation. By connecting weather information, risk locations, video analysis results, and business systems, the platform can automate multistep analysis and assessment. Agents can also coordinate work that previously required repeated manual searches and the organization and consolidation of information across multiple systems.
Urban flood prevention can therefore move beyond detecting and responding to individual events toward an integrated workflow spanning visual perception, risk analysis, agent orchestration, and coordinated task execution, turning vast visual datasets into actionable support for urban risk management and decision-making.
From Technology Leadership to Customer Value: SenseTime’s Long-Term Strength in Visual AI
According to SenseTime’s 2026 interim results announcement, its visual AI business operates in more than 20 countries and regions worldwide and has cumulatively served more than 4,500 customers. Existing customers contributed 67% of revenue, with these long-standing relationships supporting steady business growth. Across Asia-Pacific, visual AI has been deployed in smart hardware, smart cities, enterprise operations, healthcare, finance, Internet services, robotics, and other fields.
To address diverse customer needs, SenseFoundry starts with real-world use cases, integrating visual AI with existing infrastructure and business systems to make fuller use of customers’ existing equipment and data while reducing redundant investment. Modular products and capabilities support deployment requirements across different stages and scenarios. Capabilities such as specialized model training and management also help customers build proprietary AI model assets, allowing their AI capabilities to evolve alongside their businesses.
For technology deployment, SenseFoundry continues to adapt the full CV 2.0 stack to China’s domestic IT ecosystem, covering core components including specialized vision models, multimodal large models, and agent infrastructure. Compatibility with domestically developed hardware infrastructure also supports customer requirements for secure, reliable deployment environments under their own control.
Looking ahead, SenseTime will continue to advance its CV 2.0 strategy and deepen the integration of visual AI with large models and AI agents. This will enable AI not only to see and understand the real world, but also to engage more deeply with business use cases and help analyze and execute complex tasks, unlocking further value from visual AI across urban governance and industry operations.












