You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm very interested in this too. My guess is that this would be very difficult to do in the general case, because it would require calibration to correlate angular distance/pixel distance within the image to real-world linear distance.
manual calibration would require calculating the distance between the edges (or other known points) within the frame, possibly separately for all lanes. I'm sure there's libraries for doing this, but I don't know where to start.
if you wanted to try to possibly automatically calibrate, in the general case, you could detect vehicles of a known length (e.g., Tesla Model 3 is 184.8″ long) then infer the relationship between pixels and feet/metres. this sounds pretty iffy though, especially since MobileNet's bounding boxes aren't likely to be quite precise enough, especially when vehicles are photographed at an angle.
Curious @nickwebcouk if you've made any progress on this (via this or another project). If not, would be happy to chat about it, as speed monitoring is part of my goal as well.
Hello - is it possible to also capture the speed of counted vehicles/cyclists?
The text was updated successfully, but these errors were encountered: