Distracted Driving Detection utilizes Samsara’s dual-facing AI dash cam (CM32), computer vision, and artificial intelligence (AI) to analyze drivers’ head positions to determine a baseline for normal driving behavior. This information is used to help improve safety by proactively identifying distracted driving behavior and sending in-cab audio alerts that encourage drivers to keep their eyes on the road.
When the dash cam detects a dip in the head position while the vehicle is moving at a speed of >25 mph, audible voice coaching alerts notify drivers to keep their eyes on the road with the following severity levels:
- low distraction detected: 2 beeps
- medium distraction detected: 2 beeps plus voice coaching
- high distraction detected: 2 beeps plus voice coaching plus 2 more beeps
- To view longer videos or the road-facing footage as well, initiate a 15-second video retrieval from the distracted driving safety event page then proceed to the Video Retrieval page
Activating Distracted Driving Detection
- Navigate to Settings > Safety > AI Event Detection.
- Toggle the AI setting On. You will know it is on when the toggle turns green.
- Under Inward Camera, toggle the Distracted Driving Detection setting on. You will know it is on when the toggle turns green.
- You can turn on In-Cab Audio Alerts for drivers so they hear a warning in real-time
Notes:
- New customers after July 1st, 2020 will have Distracted Driving automatically turned on for their organization.
- Given the high volume of events (especially for speed threshold of 0 mph and 10 mph), there is a maximum of 3 events uploaded to the Inbox per hour per driver. The in-cab audio alerts will continue to play in cab whenever the trigger thresholds are met, even if outside of this cap.
Speed & Severity
These settings will be available under the Distracted Driving toggle on the Cameras & Safety Settings page. Changing these settings will impact the conditions that must be met before an in-cab alert is triggered and a Safety Inbox event is uploaded.
-
Speed:
-
What is it? The speed that the vehicle must be driving before an in-cab alert is triggered and a Safety Inbox event is uploaded.
-
Options: 0 mph, 10 mph, 25 mph [default], 35 mph
-
Details:
-
Increasing the speed threshold will reduce the volume of events.
-
Decreasing the speed threshold will increase the volume of events.
-
Given the high volume of events (especially for 0 mph and 10 mph), there is a maximum of 3 events uploaded to the Inbox per hour per driver.
-
-
-
Severity:
-
What is it? Severity corresponds to how long it takes a driver to go from a “distracted” state to a “normal” state. The in-cab alert starts firing when a driver has not returned their full attention to the road (recovered) within a fixed time threshold.
-
Options:
-
Low (all events): This option will generate the most number of events.
-
Medium (medium & high severity events) [default]: Events with a medium and high severity with be uploaded and shown in your Safety Inbox.
-
High (high severity events only): Events with a high severity with be uploaded and shown in your Safety Inbox. This option will generate the least number of events.
- Increasing the severity will reduce the volume of Distracted Driving events.
-
-
Safety Score Impact
Distracted Driving will impact the driver’s safety score as a duration, meaning the % of drive time that the driver spends “distracted” will be calculated and go against their score.
Duration is used (rather than the count of distracted events) given the very high frequency of distracted driving events. Distracted Driving will be one component of the AI Behavior noted in the Safety Score formula:
The duration corresponding to Distracted Driving events that are dismissed will be removed from the score, and will not impact the driver’s safety score. Administrators can also edit the weight of distracted driving using the configuration options on the Safety Settings page.
Reviewing Distracted Driving Events
Once Distracted Driving Detection is triggered, you will automatically receive a Safety Inbox event with the label Distracted with a 5-second inward-facing video.
If the event was a false positive, you can Dismiss the alert. You can also remove the Distracted label by clicking the X next to it or add labels to the event by clicking +Add.
Known Limitations
Distracted Driving Detection can sometimes trigger false positives or inaccurate alerts. The following list includes some known limitations of our detector:
- Detector can lose track of drivers' faces and doesn't trigger an event when it should
- Hand covering mouth or face may confuse the detector
- Head "dipping" or lowering while looking ahead or looking side to side can cause a false positive
- Glasses, hats, beards, masks, and objects hanging in view of the camera may cause accuracy issues
- Remove objects that may block the camera's view of the driver's head
- Detection and alerts are less accurate in extremely dark or bright conditions, such as at night or when there is a glare from direct sunlight
- Distracted Driving Detection does not detect mobile phones.
- This means that the cameras will not detect distraction when a driver is talking on the phone and looking ahead at the road.
- Distracted Driving Detection does not detect drowsy driving.
- This means that the cameras will not detect distraction when a driver yawns or closes their eyes.
- Note: The "alert" badge on the video indicates that a DDD alert was detected - regardless if an in-cab alert went off.
- Note: The in-cab alert may not be captured in the 5 sec video. The camera has to detect 5 or 6 or 7 seconds (depending on the severity setting) of on and off distraction before the event passes the threshold and therefore generates an in-cab alert. The video upload, however, is centered around the moment of distraction so that the customer can view the distraction event at it's midpoint (when it's likely the most interesting).
-
Fleet managers can initiate a 15 second video retrieval from the DDD safety event page then proceed to the Video Retrieval page to view the inward-facing and road-facing videos.
Comments
0 comments
Article is closed for comments.