To manage privacy concerns, Samsara can blur most uploaded and saved footage for all individuals (drivers, passengers, and pedestrians) and license plates. Blurring can be applied to videos in the dashboard (safety event videos and video retrievals), downloaded videos, and trip stills.
There are three different Identity Blurring features that you can enable separately or together through the Feature Management page:
Exterior Identity Blurring (CM31 and CM32)
Driver Identity Blurring (CM32)
HD NVR Blurring (HD NVR system)
As soon as you enable your preferred Identity Blurring features, Samsara applies the settings to any future image and video captures.
Samsara does not retroactively blur images and videos captured before the feature is enabled.
Identify Blurring also has the following exceptions and usage notes:
If you have Live streaming enabled, using Take New Still will capture unblurred images and videos, however, only users with explicit live streaming permissions can initiate live streams.
In the event that you need to retrieve unblurred footage such as to exonerate a driver or as requested by law enforcement, Dashboard admins can download unblurred images and videos. For more information, see Download Unblurred Footage.
Limitations of our Blurring Algorithm
Our blurring algorithm is based on AI object detection models of human heads and vehicles as “objects.” The following may degrade our blurring algorithm’s performance in identifying human heads and vehicle objects:
- Object Size: human head and vehicle objects that are too close or far from the camera might not be detected.
- Object Variation: human face and vehicle objects are broad object categories. Inherent intra-object variation of such objects may cause incorrect detections. Similarly, images with multiple human heads and vehicles might also be incorrectly detected.
- Image Orientation: Needs visible “object landmark identifiers” to work correctly. E.g., for human heads: identifiers to locate a human head, such as shape, size, hair, eyes, nose, mouth, or glasses. Please note: our blurring algorithm does not utilize human head landmarks to perform any facial recognition. It relies solely on object detection. Whereas for vehicles: identifiers such as headlamps, car paint, and car placement relative to a road. Objects that are at an angle to the camera might not be detected because all or part of their landmark identifiers might be missing.
- Object Location: objects found in uncommon image locations e.g., a human head not situated within the usual seat location e.g., a child’s face, may not have precise detections.
- Illumination and Lighting: poorly or overly illuminated objects might not be detected (e.g. bright daylight, evening/night time, and glare).
- Occlusion: partially hidden or obstructed objects might not be detected.
- Blur: blurry objects might not be detected. The quality and resolution of the relevant image may also degrade performance.
- Motion (in video): rapid movement between frames might degrade object detection accuracy.
Please sign in to leave a comment.