installation is fantastically simple and IoT is magic
I struggled a bit to find the good area to configure. And when I tested again, the website said I had overpassed the number of views. Maybe add more images that can be generated in the first couple of days.
you could also use the sound in connection to the image as a signature of vehicles…
Then a few issues:
detection does not work well in low-light conditions and even in medium light conditions.
really no way to distinguish motobikes / mopeds + scooters / bicycles?
Thanks @loicbertrand - it is so wonderful to finally start to get customer feedback on the S2 and hopefully we’ll hear from many more in the next few weeks and months.
We’re also glad to hear the installation went smoothly - as this is one of the key drivers of the development.
In answer to your comments:
is this about setting the Region of Interest (ROI)? Yes, this is limited to avoid any issues, but it should reset after a while. The key to setting the ROI is to select a segment that will show just the relevant lanes of different street users you want to count. A narrower focus helps to avoid extraneous objects (e.g. birds) being tracked and then ignored.
That is interesting. The light conditions will affect the count to some extent for contrast. The device has a good ability to deal with different light conditions so should be fine. What is your observation based on? Give it a few days to get more data and then maybe we can discuss it again?
We are hoping that some users will be able to do some validation counts with us to help test the device in different conditions. I’ll share a post on that process in the near future (no rush - we want to have a few more devices in the field first).
This is coming in the dashboard. The S2 is differentiating between pedal cycles and motorbikes, but we are not yet showing this data separately in the dashboard. Once we have more data from more devices we can update the API to reflect this, and then we will also update the device dashboards to split out the data.
The camera is able to count in many light conditions, but we are still refining the “night count” feature, so this is not currently active so you’ll see some specific cut offs in the counts.
We hope to have more data and details on this soon
I only set up my S2 yesterday, but I can confirm the low light issues. You can see that basically the counts were zero between 20:00 and 06:00 (roughly sunset to sunrise) Telraam | North Allen Street
The camera is able to count in many light conditions, but we are still refining the “night count” feature, so this is not currently active so you’ll see some specific cut offs in the counts.
So you’re saying that you intentionally don’t include night-time hours in the counts?
I’d be happy to help with validation count efforts!