Videoloft’s Smart Motion model
Our Smart Motion feature is designed to detect and track people and moving vehicles, to help you find the footage that matters faster.
Video is uploaded to the Videoloft cloud when motion is detected, or continuously if 24/7 recording is enabled.
For each second of uploaded video, five still frames are analyzed using a person and vehicle detection model.
Each detected person or vehicle is tracked as it moves through the camera view, allowing Smart Motion to follow the same object across the scene and understand its path of movement.
These detections can also trigger alerts and appear in the Smart Motion history, where you can search and filter events by person or vehicle, within specific zones or across line crossings.
Accuracy
Smart Motion accuracy depends on two key factors:
Camera position and maintenance
For best results, people and vehicles should be clearly visible within the camera view. For outdoor infrared cameras, offset IR lighting can improve nighttime detection by reducing insect activity around the camera.
AI model performance
Our Smart Motion model is continuously improving, but some detection limitations remain. The section below outlines each model version, along with its improvements and known limitations.
Model releases
Each release is tested extensively and benchmarked. Below you’ll find version history, accuracy scores, and known limitations.
V3 Improved detection in low light and partially obscured scenes Oct 2026
Release Date October 2026
Summary New detection model improves accuracy and reduces false alerts.
Smart Motion V3 introduces a new detection model that identifies people and vehicles more accurately, particularly in low light or when they are partially hidden behind objects such as walls, bushes or parked vehicles.
This improves detection in challenging conditions, reducing missed events and false alerts.
Example
The example images below are AI-generated to protect customer data.
More accurate detection of people and vehicles in dark or poorly lit areas.
Better tracking of people and vehicles, even when partially hidden by obstructions.
V2 Improved detection near occlusions Aug 2026
Release Date August 2026
Summary Improved detection of people and vehicles moving near obstructions.
Smart Motion V2 improves detection when people or vehicles move into or out of a detection zone from behind an obstruction, such as a bush, wall or parked vehicle.
Previously, these movements could be missed if an object became visible only after entering the detection zone, or disappeared shortly after leaving it.
Example
The example image below is AI-generated to protect customer data.
People and vehicles moving into or out of detection zones near obstructions are now detected more reliably, reducing missed events in partially obscured scenes.
V1 Initial release Jan 2026
Release Date January 2026
Summary First production release of the Smart Motion model.
The first release of Smart Motion combines a pre-trained, industry-standard person and vehicle detection model with Videoloft’s own bespoke tracking technology to deliver searchable, deduplicated results.
V1’s engineering focus has been on building robust tracking – linking detections across frames, reducing duplicate events, and enabling accurate searching by area and line crossings. The pre-trained model has some limitations in detection accuracy in certain scenarios, which are outlined below.
V2, which is already in the works, addresses these limitations. We are retraining the person and vehicle detection model using real-world CCTV footage from thousands of opted in Videoloft customer cameras, just like we did for our ALPR, People Counting and PPE detection models. This will improve detection accuracy across a wider range of environments and conditions, improving the accuracy of Smart Motion results.
Known limitations
The example images below are AI-generated to protect customer data.
False detections
Objects can be misclassified as people/vehicles, particularly in low-light or infrared night vision.
A stationary vehicle can be incorrectly identified as moving when another vehicle passes through the scene.
If a person is fully occluded for more than 2 seconds, they may be detected more than once when they reappear.
Strong or well-defined shadows may be incorrectly detected as an additional person.
A person carrying a wheel-like object may be incorrectly classified as a moving vehicle.
Animals may be misclassified as people, particularly when they are distant or partially obscured.
Missed detections
Vehicles may not be detected in low-light infrared scenes where headlight glare obscures vehicle features.
hello@videoloft.net
Toll-free US: (855) 737-0237
UK: +44 203 9870133
+44 753 280 1187
Try it for free
For more information, contact us directly or fill in your details using our quick form and a representative of our global sales team will be in touch.
Your details
We value your privacy and take data protection seriously. We will never share or sell your personal data to third parties. Your trust is important to us, and we’re committed to safeguarding your information.