SmartVision 6.6 Delivers Faster Face Recognition and New AI Detectors
SmartVision 6.6 is now available, bringing faster face recognition, lower GPU usage, new AI-powered detectors, and improved multi-monitor support.
The biggest change in version 6.6 is a completely redesigned face detection and recognition model. The new model runs several times faster than the previous one while using fewer GPU resources. In practical terms, this means the same computer can now process more camera streams with face recognition enabled.
The SmartVision 6.6 trial is now available for download from the SmartVision website.
Faster Face Recognition
In SmartVision 6.6, the model responsible for face detection and recognition has been completely replaced.
The new model is several times faster and requires fewer GPU resources. This is especially important in systems with many cameras, where GPU utilization directly affects how many video streams can be analyzed at the same time.
The same hardware can now process more camera streams with face recognition enabled, reducing the need to upgrade to a more powerful GPU as the system grows.
Users can also adjust the face matching threshold in the INI configuration file. By default, the trial version is configured with a relatively low matching threshold, allowing the system to return visually similar faces in addition to close matches.
For applications that require stricter identification, the threshold can be increased in the settings.
More Cameras on the Same Hardware
The new face recognition model is the first step in a broader optimization effort.
In future SmartVision releases, we plan to apply the same approach to other neural network models. The goal is straightforward: allow the same hardware to support more cameras and run more AI analytics at the same time.
SmartVision is evolving not only by adding new features, but also by making more efficient use of existing computing resources.
New AI-Powered Camera Tampering Detection
SmartVision 6.6 also introduces a new AI-powered camera tampering detector.
The system can detect situations where a camera is still online but is no longer monitoring the intended area properly.
For example, SmartVision can detect when a camera lens is covered, when the camera is suddenly moved or turned, or when its field of view changes abruptly.
This helps identify incidents that a basic camera health check may miss. The camera may still be transmitting video, but the actual scene being monitored has changed or become unusable.
When tampering is detected, SmartVision can automatically generate an event and alert the operator.
Smoking Detection
Version 6.6 also adds a new smoking detector for locations where smoking is prohibited.
Instead of relying on a single detection model, SmartVision uses a multi-stage analysis pipeline.
The system first detects a person, then estimates pose, tracks the person, analyzes the hand region, and looks for signs of a cigarette.
This makes it possible to analyze the context of the scene rather than simply searching the entire image for a very small object.
The feature can be used in industrial facilities, warehouses, transportation facilities, public buildings, and other areas where smoking is restricted for safety or operational reasons.
When a smoking event is detected, SmartVision can alert the operator or trigger a predefined response.
Weapon Detection
SmartVision 6.6 also introduces a new AI-based weapon detector.
If the system detects an object that matches the configured criteria, it can automatically generate an event, alert the operator, and bring the relevant camera feed to their attention.
The weapon detector can be used together with person detection, object tracking, and SmartVision’s event system.
This is particularly useful in environments where operators cannot continuously monitor dozens or hundreds of live camera feeds.
Improved Multi-Monitor Support
SmartVision has supported multi-monitor setups for some time, but version 6.6 makes them more convenient to use.
The software can now remember the position and layout of its windows across multiple monitors.
For example, an operator can place the main camera grid on one display, events and analytics on a second, and an enlarged camera view on a third.
When SmartVision is closed, the workspace layout is saved. The next time the application starts, the windows are automatically restored to their previous positions.
For security control rooms and monitoring stations, this eliminates the need to rebuild the workspace after every restart.
Vehicle Analytics Continues to Expand
Previous SmartVision releases also introduced emergency and special vehicle detection.
Combined with license plate recognition, this makes it possible to create more flexible vehicle access and event-processing scenarios.
For example, SmartVision can take into account not only a vehicle’s license plate, but also its detected category, and then trigger different actions depending on the analysis results.
This allows the system to move beyond basic license plate recognition toward more comprehensive vehicle analytics.
More AI Analytics Without Constant Hardware Upgrades
SmartVision 6.6 reflects a broader development goal: getting more analytics from the same hardware.
Adding new AI models only makes sense if they can be used efficiently. A new feature should not automatically mean a new computer or a more powerful GPU.
The redesigned face recognition model is the first major step in this direction. Future releases will extend the same optimization approach to other SmartVision AI models.
The SmartVision 6.6 trial is now available for download from the SmartVision website.