You have two recording schedules, person detection, and ring.
If the doorbell rings, it should not duplicate video clips with object detection, you will have a Button push triggered video clip.
So that we can be of assistance in situations like this, please always be sure to include specifics such as the time that the event occurred, the time-stamp of any video clips you are referencing.
I see a button push video event recorded in your list for today at 11:20. Is that the event you are referencing?
Yes, that is the one I was talking about. I received a notification of a button press, but was out, so missed it in real time, but wanted to check who had tried to reach me.
I looked in the app highlights timeline: no photo or video. I looked in the app notification history, no photo or video.
Only an hour or so later, long after I wrote you, did the video show up, with unhelpful backdated timestamp.
I was asking: where is the video? And I guess the answer is probably: it can sometimes take a long time for the server to process the snippet, and for real time use when we get a notification without video or photo, just click the bell placeholder to start streaming instead.
And for my use case, just wait as long as it takes for the snippet to be made available.
Is that right?
PS
Per your idea about redundant person detect vs doorbell, my thought is that I want a video or photo immediately in either case. Person detect photo notifications have so far been delivered within 30secs or so, which is ok. If doorbell will always take as long as it took today, then I’d rather have both working at the same time. Rationale: I want “caller ID”. I don’t want to open a two way stream to find out who it is, accidentally engaging some random sales person.
Even the person detect feature I find pretty confusing. Sometimes notifs are fifteen secs, sometimes a few minutes. If I click a push notification from person detect most often it opens a stream on nothing, as they are long gone. And even the photos saved are sometimes useless. For example:
My Surety System: Front doorbell detected a person at 4:31pm
Another thing that’s puzzling when trying to debug this is that the ADC app rewrites reality with it’s notification history feature.
When I get a doorbell notif, without image, hit the button, see no one, then go outside, then come back in, then later get video of the push event, this is not the way we want this to work.
Stranger still: the ADC in-app notifications history rearranges this sequence to reflect the order we wish we had: 1. push, 2. video, 3. door open.
Yet Android notifications history shows that the notif with video of the push actually arrives several minutes after the other events.
So if there’s no way to speed up these video snippets taking so long to arrive, is there any way to have a still image attached to the very first button press notif?
To clarify expectations of what is going to happen, recordings of any kind will not be available while the recording is happening. It will take a few minutes for these to show up in the list. (It shouldn’t be a long time, but it will always be after the recording is completed.)
Only an hour or so later, long after I wrote you, did the video show up, with unhelpful backdated timestamp.
It is possible for a delay in odd occurrences, if there is a network connection or app issue., but it shouldn’t be this long. Were you refreshing the app that whole time, for an hour?
Person detection with the thumbnail attached should go out right away after the thumbnail is processed by ADC, not the whole recorded clip. That notification should link you to live view since the clip is not ready yet when the notification is sent. The person that triggered it may no longer be around.
Button press notifications are sent right away, and link you to live view. Recordings should be available within a few minutes typically.
So if there’s no way to speed up these video snippets taking so long to arrive, is there any way to have a still image attached to the very first button press notif?
We can push this to ADC, I think if there is any impact on processing time they may not, but if it can be implemented it is a good idea.
And even the photos saved are sometimes useless.
You may need to adjust the ground zone for that rule, and potentially sensitivity if reflections or shadows are being caught.
Can you send us a private message with a screenshot of the ground zone selected within the rule?
“Were you refreshing the app that whole time, for an hour?”
Ha! Yes I can see that might cause some problems. No, just observing the Android notification time stamp.
“Button press notifications are sent right away, and link you to live view.”
Yes but they go out without an image. It’s a camera doorbell. They should send a thumbnail with the alert.
“We can push this to ADC, I think if there is any impact on processing time they may not, but if it can be implemented it is a good idea.”
Yes makes sense they might worry about that, but on the other hand their product is not as advertised and not competitive as is, so I hope they would take that into consideration too. It falls laughably short of the cool video that sold it to me. It would be awesome if you would offer that suggestion.
“Can you send us a private message with a screenshot of the ground zone selected within the rule?”
I did not set up a ground zone (the whole field of view is useful in my mounting location). The bug is in the software which selects which frame to share as a thumbnail, see the shot I shared before.
I did not set up a ground zone (the whole field of view is useful in my mounting location). The bug is in the software which selects which frame to share as a thumbnail, see the shot I shared before.
A ground zone or trip wire is part of any video analytics rule setup. If you let the whole frame be set as a ground zone this can cause problems with video analytics if only a portion of the image is walkable ground.
In your image, only part of the left side of the screen would be ground zone.
The “ground zone” is literally the ground area where individuals would walk. This allows the camera to process what is human movement vs other motion, and if the whole frame is selected, or area outside of the walkway, that would be a reason why you would see non-useful video recordings labeled as persons. Please be sure to edit that ground zone accordingly for best results of the person detection rule.
Send us a screenshot of that rule page with the ground zone if you have questions.
It falls laughably short of the cool video that sold it to me. It would be awesome if you would offer that suggestion.
Always happy to forward suggestions, and again I agree I think it would be a good addition.
I originally double checked in ADC documentation for reference to this because at first I thought it was already an option. I’ll let you know if they provide some info in follow up to the request.
I thought ground zone and trip wire were same thing, and used for example to remove a sidewalk area from a front camera so only people in the area of interest are flagged.
The system did not require that I set one, and I didn’t because my camera can only see the doorway area, and I’d like to know if a person is ever in the field of view.
But if a ground zone and an exclusion zone are not the same thing, and you actually need to somehow outline the ground plane in every camera view, that seems unfortunate, it should be smarter than that. And also if it’s required, then the system should require it.
I’ll go back in and add one today.
Yet, I hope you agree that the image I shared is an obvious bug. That is not a person. And examining the clip, there was a person, but the thumbnail was just the wrong frame.
And if this is not just a simple aliasing bug, but because there is no “ground zone,” then their person algorithms need some serious work!
That is not a person. And examining the clip, there was a person, but the thumbnail was just the wrong frame.
I can send the info to ADC and open a ticket, but the first thing they will recommend is to edit the ground zone, so we should test that first.
The thumbnail being of the wrong frame is likely the first (I would assume, this is not documented) ‘person’ detection frame from the clip. It’s certainly odd, but the analytics was clearly looking at something much smaller than a human in that space without the defined zone. Setting the ground zone should help a good deal.
Question on this: while searching for where to set this up, I noticed that web app and mobile app are extremely similar UIs, except that the web interface shows me most recent views from my doorbell and my other camera, but the mobile app always just shows a blur and requires starting video stream to see anything.
Is there a way to get the mobile app to offer the same “latest image” for each camera that the web app version does?
Additional question: I went in and set a ground zone according to the instructions, but it didn’t make sense to me, because people’s feet are not visible in the camera view, the camera sees their upper body. Can you help me understand this better?
Is there a way to get the mobile app to offer the same “latest image” for each camera that the web app version does?
The app will show a still image from recent activity for that cam. In my app both of mine do currently, a V515 and V724.
Is that the iOS iPhone Alarm.com app?
Additional question: I went in and set a ground zone according to the instructions, but it didn’t make sense to me, because people’s feet are not visible in the camera view, the camera sees their upper body. Can you help me understand this better?
Feet don’t need to be in the frame, the ground zone is the general area where humans will be able to walk through, helping analytics determine where human motion will occur/expected size etc. The ground zone should just be the visible ground area in the frame, avoid walls and obstructions where people won’t be.
Sometimes it is important to set the ground zone smaller to avoid people walking down the sidewalk, for example, but in your frame you would just want to filter out the non-ground area, mostly on the right.
“The app will show a still image from recent activity for that cam. In my app both of mine do currently, a V515 and V724.”
For me, the mobile app main screen shows the last person detect event thumbnail for the doorbell, while the web app shows the very latest view from the camera, regardless of AI, which is what we want. The app should also do this.
Likewise with my interior camera, the app shows just a blur instead of a most recent image, while the web app shows the most recent image as expected.
The attached screenshots of web app and mobile app are all from the same time so should all show the same images.
“Is that the iOS iPhone Alarm.com app?”
No sorry Android.
I set the ground zone to cover just the slate walk seen by the doorbell camera, cropping out walls, door and garden. Yet the system still creates spurious person detect events (notifications and video snips without persons). Most recent example is timestamped 4:55 pm 4/10/22. So I think the bug is not ground zone related.
Likewise with my interior camera, the app shows just a blur instead of a most recent image, while the web app shows the most recent image as expected.
If you have analytics recordings for the camera, then you should see a still of the clip. I see that on my own V515.
I don’t actually see a video analytics recording rule for that camera on your account. You can set one up in Video > Recording Rules.
For me, the mobile app main screen shows the last person detect event thumbnail for the doorbell, while the web app shows the very latest view from the camera, regardless of AI, which is what we want. The app should also do this.
Happy to send this to ADC as a request.
I set the ground zone to cover just the slate walk seen by the doorbell camera, cropping out walls, door and garden. Yet the system still creates spurious person detect events (notifications and video snips without persons). Most recent example is timestamped 4:55 pm 4/10/22. So I think the bug is not ground zone related.
We can’t actually view the video clips, can you send a private message with a still of the person detection the rule captured?
“If you have analytics recordings for the camera, then you should see a still of the clip. I see that on my own V515.”
Basically I am just comparing and contrasting the features of the mobile app vs web app, where I can see latest thumbnails regardless of rules.
I think if you are willing to pass on my suggestion that the mobile app fully duplicate the web app and vice versa this would solve it.
Right now, the web app tells us we must go to the mobile app for some things, and the mobile app tells us we must go to the web app for other things, I just wish there were only one place to administer.
I think there are at least four apps and urls across which various account settings, system options, system configuration things are set: adc mobile app, adc web app, adc account pages, surety account pages…
“We can’t actually view the video clips, can you send a private message with a still of the person detection the rule captured?”
Ok, I think I included a screenshot on the thread above, there is no person. I’ll make a short video on my phone and pm you the link.
Go for it, thank you for the help. Also you might include the screenshot I posted in this thread of another event which outlined a spot on the door as a person.
