Hi
I’m running two devices at one address on the same window.
V1/Raspberry id 9000010984
S2 id 9000011401
The S2 will be moved down the street when I’ve found a nice neighbor who is willing to install it at his home.
When comparing the statistics of both, I see big differences. Counters and speed.
Has anyone had any experience with this?
tnx
Pascal
Hi Pascal, and thanks for joining us here.
Indeed, the data captured by the V1 compared to the S2 is quite different. The V1 was a proof of concept design that had a simplistic algorithm for identifying objects. It was ‘best guess’ from the shape of the object, and required some local calibration.
The S2 works in a very different way, “seeing” and recognising different objects, and is therefore likely to be a lot more reliable.
I can’t comment on which data is closer to the ‘ground truth’ on the road (I assume they are on the same segment, despite being on different ones on our map?). If you could carry out a manual count for 15 or 30 minutes, we can compare this to the data of the two devices and see where there is a discrepancy. I see a big difference in “large vehicles” between them which is something that can happen on residential roads with the V1 as it self-calibrates to try to identify large vehicles relative to the majority of vehicles.
I did check the ROI on your S2 device and it appeared to be set a little too wide, so I have narrowed it slightly to help see more of the smaller objects at that distance. However, I also note from your data that there are sometimes strange data patterns. On some nights there is no traffic at all, whilst on others it can be very busy. Is this a pattern you recognise? Alternatively, is it a case that you close external shutters at night normally (so collect no data), and on occasions they are left open and collect the data?
Hi Rob
Hi
Before hearing from Telraam (which is a great project btw) I had been running a Github project.
https://github.com/pageauc/speed-camera on my raspberry pi. When comparing speed statistics with the Telraam V1 device, I think they are more or less equal.
Correct, both devices are physically located at the same address and in the same window. The S2 is just pointing to a different road segment. I’m still looking for a kind neighbor who is willing to let me use their window.
It took me a while to count the vehicles manually. From my point of view, I consider a Mercedes Sprinter to be a large vehicle. Is that classification okay?
You’re right, I didn’t notice that at night. The shutters are usually closed during the night. There are also some LEDs that blink quite a lot. They might be reflecting in the window and could be causing that effect.
Here are my countings
| Date | 01.07.2026 | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| TIME | 12:00 – 13:00 | 13:00 – 14:00 | 14:00 – 15:00 | ||||||
| Security cam | V1 | S2 | Security cam | V1 | S2 | Security cam | V1 | S2 | |
| Cars | 17 | 14 | 12 | 25 | 17 | 15 | 30 | 34 | shutters too low |
| Heavy vehicles | 7 | 7 | 0 | 7 | 10 | 0 | 8 | 19 | shutters too low |
actually, probably not. This would be a “Light truck” category (which includes vans) and therefore be included in the Heavy Vehicles general total.
I am however surprised that the S2 is not registering more heavy vehicles and generally undercounting. I can’t think why this would be and it could be the LEDs, or maybe something else. I have sent an update for the firmware to your device to see if this helps. Another option would be to simply reposition it in this window or a different one to see if it improves accuracy with a better / different view.
As for the shutters, that explains the data patterns which appear to be rather unusual. It would really help to have this in a window that does not need to have shutters so that it captures all hours of the day.