|
|
| In 2025, the video surveillance industry is witnessing a transformative wave, driven primarily by the rapid adoption of vision language models (VLMs). As generative AI and large language models like ChatGPT gain traction worldwide, the industry has pivoted to models that combine vision and language. This shift enables systems to understand and retrieve video content in ways that are more aligned with human semantics, opening new possibilities for intelligent surveillance applications. |
| |
Custom AI Models: From Off-the-Shelf to In-House Innovation
Jason, the platform manager at Network Optix (Nx), highlights the evolving needs of end users and integrators, who are no longer satisfied with generic, pre-trained models. With the increasing diversity of use cases, organizations are now forming their own AI teams, collecting proprietary data, and training models tailored to their unique operational requirements. This push towards customized AI enhances accuracy and delivers greater business value.
The Proliferation of Acceleration Chipsets
The landscape of AI hardware is expanding rapidly. While established names like Nvidia and Intel remain central, a wave of startups—including Hailo, DeepX, Axelera, EdgeCortix, and SiMa—are injecting more choices into the market. Users now have a wider range of inference platforms to match specific workloads, which in turn challenges software providers to keep up with compatibility and optimization demands across a broader array of devices.
Despite advances in AI model training, Jason points out that the most difficult part of innovation is the deployment stage. Scaling from a proof-of-concept to an operational environment—such as deploying AI models across hundreds of cameras in a factory—requires seamless integration and performance tuning. As hardware diversity increases, the effort needed to ensure software and hardware work together efficiently grows even greater.
Video surveillance platforms now aggregate not only video, but also data from LiDAR, radar, and a variety of IoT sensors. Presenting this influx of information in an intuitive way has become vital to maximizing the value of these systems. Simultaneously, the transition from edge to cloud architectures has heightened the industry’s cybersecurity standards, necessitating new strategies to safeguard sensitive data.
Nx EVOS: Meeting Challenges with Innovation
Nx AI Manager: One-Click AI Deployment from Cloud to Edge
Nx EVOS addresses deployment complexity with its new Nx AI Manager, allowing users to upload, select hardware, and deploy AI models with a single click. The platform supports model conversion and optimization for leading chipsets and toolkits—including Intel OpenVINO, NVIDIA CUDA, Qualcomm QNN, and Hailo RT—empowering users to unlock the full potential of their chosen NPUs or GPUs with minimal technical friction.
Nx Maps: GIS Visualization for Video and IoT Data
Nx Maps responds to the challenge of data visualization by providing an integrated 2D/3D GIS interface. Video streams, AI-detected objects, and sensor data are layered onto interactive maps, offering users a comprehensive overview of their facilities and enabling faster, more informed responses to emerging situations.
Security and Compliance: SOC 2 Type 2 and ISO 27001
Recognizing the critical role of cybersecurity, Network Optix has achieved SOC 2 Type 2 and ISO 27001 certification, ensuring not just the security of its software but also the rigor of its organizational processes. These standards guarantee that both data protection and operational reliability remain at the core of the company’s offerings.
Through robust AI management, next-generation mapping, and a deep commitment to cybersecurity, Nx EVOS is equipping users to accelerate innovation and secure a leading edge in the rapidly evolving world of AI-powered video surveillance. |
| |
| |
※The text and images in this article may not be reproduced without authorization. For licensing inquiries, please email contact@aimag.tw — [iDS Magazine Statement]※ |
|
|
|
| |
|
| |
|
