I wanted to know when our dog was stood waiting patiently at the back door to go out, so I set a zone and filtered out humans ;-) All from following your first two Frigate videos 👍
Glad I was able to help. That's an interesting use case. When I let me dog out, I pull of the camera to see when she is ready to come in. It would make more sense to do what you are doing so I don't have to stare at the screen.
Solid video Chris! I need more Frigate in my life! I have it setup and also turned off Blue Iris for the moment. Using it to detect my dog coming in and out of the back door. it's alittle slow for notifications. Working on speeding it up. I don't have a Google Coral.
Frigate has been working great for me. As for the notifications, I am using some stuff for that and will show it off in the next video. What are you using to get notifications and what device is it slow on?
@@mostlychris I have a unRAID server 48 core xenon. And I listed each cpu core as a decoder. Not sure if that config is correct or not. I'm using a home assistant blueprint for notifications. Found It on the blueprint exchange it works really well
That's some beefy hardware. It sounds like notifications are not related to processor but maybe delay between detect and the notification firing off. I'm on an Android device and I added a couple of lines to make the HA helper app have a higher priority for frigate notifications.
I was going to put this in my latest but I forgot. It's going to be hard to see this in youtube comments, but I added ttl:0 and priority:high to each action in the frigate notification blueprint. action: - choose: - conditions: '{{ not group_target }}' sequence: - device_id: !input 'notify_device' domain: mobile_app type: notify message: A {{ label }} was detected on the {{ camera_name }} camera. data: tag: '{{ id }}' group: frigate-notification-{{ camera }} image: /api/frigate/notifications/{{id}}/thumbnail.jpg?format=android attachment: url: /api/frigate/notifications/{{id}}/thumbnail.jpg ttl: 0 priority: high
There is some useful info here that I think I can use for alerts. I masked off areas on each of my 4 cameras where trees are moving and such and it didn't actually have much affect on CPU, so not sure if masks will help me much here. I'm using a coral USB device and getting 65-70% CPU utilisation with or without the masks.
Great series of videos, thanks! 😀. Question please, if you had a zone set up for a person, would an alert be triggered when just a leg enters the zone or the entire human enters it? If entire person, wouldn't you want the zone bigger? I know you were just given an example but on your back porch, you probably don't get a persons head, within the zone outline much. Cheers
The description in the docs say it looks at the bottom of the box. In addition, it looks at multiple frames to determine an object. I've seen people triggered across the street when their head is out of frame and only legs are visible. Of course, trial and error (fine tuning) is recommend for each camera in each environment as there will be subtle differences.
In you final thoughts you tell about using the substream. Should I use always the substreams for Frigate? I have a 5MP cam at the backyard and wonder how I can save a little bit CPU time since HASSIO and Frigate run on a 916+ Synology. What would be the best way to go?
In all that I have read, using the substream will save on CPU cycles. That is why I use it. I have 5 cameras running 24x7 on my Odroid and use the sub stream for detection and the main stream for recording. That way I get the good quality recordings.
I'm doing exactly that right now. I've turned off blue iris for now. Still like BI but Frigate serves my use case at the moment. Mainly it detects objects better than BI with Deepstack in my environment.
With the masking of the time/date, does this totally blank this in the recordings as well, or is the time and date still visible in recordings? I feel this may be important if you needed the feed from the camera for legal purposes.
Sure. I started moving stuff to github because the haste and paste bins are all deleting stuff or putting it behind logins. github.com/mostlychris/homeassistant/blob/master/frigate.yml
what also needs to be noted that you change the detect resolution in the config.yaml after you create the zones ... everything in zones and masks goes bonkers ! it won`t re-adjust to the size.
That is true. The zone coordinates are based on what would be a static view of the camera image. If you change the resolution, the camera image changes size and the coordinates don't follow so you'll need to adjust the zones after changing resolution.
@@mostlychris I kept playing with the streams and the resolution on my hikvisions because I noticed to many false positives from Double Take and deepstack hence I noticed that I had to redo the masks and zones again once I decided to stick with a resolution. In the end I decided to decode 1280x720 to be honest to get more pixels in the snapshots, I also ditched deepstack and went with compreface quite a difference. Obviously now I'm playing with room-presence detection via Double take which is more awesome that I thought. I have some light automations that now I can simply turn on and off based on presence and the nice thing is the last known presence doesn't change in the sensor until new camera detects you which is really powerful! Now I'm going to ditch my ring indor cams and get some rtsp ones. Which would you recommend ?
I'm playing with Reolink right now and will have a review video at some point. I don't have much in the way of indoor stuff (family reasons) but there are probably people on here that can suggest.
When I begin clicking on the image to create a mask, the first point is way below the image on the screen. So confused lol. Any idea what would cause this?
@@mostlychris I was able to resolve it by removing the previous masks that were in the config file. I am not sure if I inadvertently put them there when setting things up or what. But all well now.
I have a Car parked on my driveway in the same spot, what is the best way to have it ignored so Frigate doesn't keep recording it or alerting me Please
Would you use stationary objects for a tv hanging on a wall or motion masks? I am getting motion and person detection on the TV and would like to ignore sections of the wall to not get notified or recording movement from the TV.
Good videos, how to store media in nas ,I followed the guides/ha_network_storage/ but its not storing in the nas frigate folder. Can u share some tips.
The best vids I have come across in regards to frigate Thank you so much :)
Thanks!
You can forget earlier quistions. It's up and running, thanks to your ferry good videos. Thanks a lot.
Excellent! I'll ignore the rest 😀
Thank you so much for all your Videos on Frigate.
You're welcome and thanks for watching!
hello chris
thanks for part 2 in the frigate-story. looking forward for the next part(s)
You're welcome. Working on the next installment.
Super helpful! Thanks.
Glad it was helpful!
I wanted to know when our dog was stood waiting patiently at the back door to go out, so I set a zone and filtered out humans ;-) All from following your first two Frigate videos 👍
Glad I was able to help. That's an interesting use case. When I let me dog out, I pull of the camera to see when she is ready to come in. It would make more sense to do what you are doing so I don't have to stare at the screen.
Awesome work! Thank you for showing in such detail!
Thanks! You're welcome!
Solid video Chris! I need more Frigate in my life! I have it setup and also turned off Blue Iris for the moment. Using it to detect my dog coming in and out of the back door. it's alittle slow for notifications. Working on speeding it up. I don't have a Google Coral.
Frigate has been working great for me. As for the notifications, I am using some stuff for that and will show it off in the next video. What are you using to get notifications and what device is it slow on?
@@mostlychris I have a unRAID server 48 core xenon. And I listed each cpu core as a decoder. Not sure if that config is correct or not. I'm using a home assistant blueprint for notifications. Found It on the blueprint exchange it works really well
That's some beefy hardware. It sounds like notifications are not related to processor but maybe delay between detect and the notification firing off. I'm on an Android device and I added a couple of lines to make the HA helper app have a higher priority for frigate notifications.
@@mostlychris Oh how did you make it have a higher priority is it in the Android App or Server Side?
I was going to put this in my latest but I forgot. It's going to be hard to see this in youtube comments, but I added ttl:0 and priority:high to each action in the frigate notification blueprint.
action:
- choose:
- conditions: '{{ not group_target }}'
sequence:
- device_id: !input 'notify_device'
domain: mobile_app
type: notify
message: A {{ label }} was detected on the {{ camera_name }} camera.
data:
tag: '{{ id }}'
group: frigate-notification-{{ camera }}
image: /api/frigate/notifications/{{id}}/thumbnail.jpg?format=android
attachment:
url: /api/frigate/notifications/{{id}}/thumbnail.jpg
ttl: 0
priority: high
Great video, thanks for the time and effort for making it.
Thanks!
There is some useful info here that I think I can use for alerts. I masked off areas on each of my 4 cameras where trees are moving and such and it didn't actually have much affect on CPU, so not sure if masks will help me much here. I'm using a coral USB device and getting 65-70% CPU utilisation with or without the masks.
Make sure you are using motion masks. If you use object masks, it still processes those areas.
Great series of videos, thanks! 😀. Question please, if you had a zone set up for a person, would an alert be triggered when just a leg enters the zone or the entire human enters it? If entire person, wouldn't you want the zone bigger? I know you were just given an example but on your back porch, you probably don't get a persons head, within the zone outline much. Cheers
The description in the docs say it looks at the bottom of the box. In addition, it looks at multiple frames to determine an object. I've seen people triggered across the street when their head is out of frame and only legs are visible. Of course, trial and error (fine tuning) is recommend for each camera in each environment as there will be subtle differences.
In you final thoughts you tell about using the substream. Should I use always the substreams for Frigate? I have a 5MP cam at the backyard and wonder how I can save a little bit CPU time since HASSIO and Frigate run on a 916+ Synology. What would be the best way to go?
In all that I have read, using the substream will save on CPU cycles. That is why I use it. I have 5 cameras running 24x7 on my Odroid and use the sub stream for detection and the main stream for recording. That way I get the good quality recordings.
Hello Chris,
Do you think Frigate can be used instead of Blue Iris? Not just for object dedction but as a NVR too.
Thank you!
I'm doing exactly that right now. I've turned off blue iris for now. Still like BI but Frigate serves my use case at the moment. Mainly it detects objects better than BI with Deepstack in my environment.
With the masking of the time/date, does this totally blank this in the recordings as well, or is the time and date still visible in recordings? I feel this may be important if you needed the feed from the camera for legal purposes.
It doesn't remove it from the recordings. It just keeps frigate from having to continuously watch it and determine if it is an object.
hi Chris what about the live view grid does it show at 30 fps or at the same 5fps?
Live view will show at the fps settings of the stream or substream your camera is set to.
Hey!, do you mind please updating the link to your YAML in the video description? Its broken. Thanks!
Sure. I started moving stuff to github because the haste and paste bins are all deleting stuff or putting it behind logins.
github.com/mostlychris/homeassistant/blob/master/frigate.yml
what also needs to be noted that you change the detect resolution in the config.yaml after you create the zones ... everything in zones and masks goes bonkers ! it won`t re-adjust to the size.
That is true. The zone coordinates are based on what would be a static view of the camera image. If you change the resolution, the camera image changes size and the coordinates don't follow so you'll need to adjust the zones after changing resolution.
@@mostlychris I kept playing with the streams and the resolution on my hikvisions because I noticed to many false positives from Double Take and deepstack hence I noticed that I had to redo the masks and zones again once I decided to stick with a resolution. In the end I decided to decode 1280x720 to be honest to get more pixels in the snapshots, I also ditched deepstack and went with compreface quite a difference.
Obviously now I'm playing with room-presence detection via Double take which is more awesome that I thought.
I have some light automations that now I can simply turn on and off based on presence and the nice thing is the last known presence doesn't change in the sensor until new camera detects you which is really powerful!
Now I'm going to ditch my ring indor cams and get some rtsp ones. Which would you recommend ?
I'm playing with Reolink right now and will have a review video at some point. I don't have much in the way of indoor stuff (family reasons) but there are probably people on here that can suggest.
@@mostlychris I've just purchased two E1 pro thinking that because they have rtsp I will be able to get them into frigate ... but I was very wrong ..
@@TheRealAnthony_real What is the issue with those? The Reolink that I just reviewed has rtsp that works in frigate.
When I begin clicking on the image to create a mask, the first point is way below the image on the screen. So confused lol. Any idea what would cause this?
I haven't seen this. Might need to zoom your browser window out.
@@mostlychris I was able to resolve it by removing the previous masks that were in the config file. I am not sure if I inadvertently put them there when setting things up or what. But all well now.
I have a Car parked on my driveway in the same spot, what is the best way to have it ignored so Frigate doesn't keep recording it or alerting me Please
Use stationary objects. There is info in the docs for that.
Would you use stationary objects for a tv hanging on a wall or motion masks? I am getting motion and person detection on the TV and would like to ignore sections of the wall to not get notified or recording movement from the TV.
Good videos, how to store media in nas ,I followed the guides/ha_network_storage/ but its not storing in the nas frigate folder. Can u share some tips.