I installed the double-take add-on and compreface. I configured the add-on to connect to my mqtt broker and to compreface successfully. I’m reading through this and I wonder if double-take can be used without frigate? Can I store an image named latest.jpg or snapshot.jpg under /media/double-take/latest for example and trigger double-take through mqtt to process it? It seems possible. Somebody has done this already?
I figured it out. Double-take provides a rest api with a recognize/upload endpoint. Example:
curl -POST --location http://192.168.1.200:3000/api/recognize/upload --form files[]=@“1.jpg”
A mqtt message is sent to the “double-take/matches/xxxxx” topic. Perfect!
Hi. I saw there is new update compreface to 1.2 on github with alot new feature and fix, it will be awesome if there some one can make hassio addon compreface for this update…
Do anyone know if CompreFace now working with Google Coral M.2 TPU?
Is there a way to toggle double-takes detection from home assistant?
I’d like for it to start detection only when I’m in geofence range to save cpu/power consumption.
@Mike90 I do this with this configurations.yaml config:
rest_command:
doorbell_face_detection:
method: GET
url: http://192.168.11.213:3001/api/recognize?url=http://192.168.11.211/image/Doorbell%3f%26decode={{decode}}&attempts={{attempts}}&camera=manual&results=best&break=true
You’ll need to update the above for your situation. See documentation: Double Take
Once the above REST call is configured, it can be used as action in any automation.
Thanks for your response, doesnt this just call for one time recognition of an image? I’d like to toggle detection on and off for lets say 5 min, a bit like the schedule feature as in the documentation.
So I’d want a switch in homeassistant that turns on or off (all) recognition from doubletake. I guess I could automate a seperate mqtt topic and then only send images when the switch is on but that seems like a bit of a patch job.
I’ve recently set up doubletake and deepstack, can anyone explain to me why I get a person being recognised but its still got a red background to the detector tile?
Same question, did you ever fix this?
Just noticed that Double Take is abandonware unfortunaly is there an alternative ?
I would like to achieve something very similar, but I’d like to use a recognised person as a condition instead of a trigger, because of the so many false positive recognitions (even with high confidence levels).
Could anyone please explain me how to extract a matched person’s name of the doubletake sensor in a value template to use it as a condition in an automation?
2 Likes
thanks!!! i actually looked for forks but didnt see that one! awesome!
Can I safely assume that the telemetry site used in the fork mentioned above is in fact, just a really really really strange coincidence in that it shares a name with a 2020 russian disinformation campaign?
Hello,
Is there a way to remove specific entities from home assistant?
I’m trying to remove an old match, but the home assistant entity has the remove button grayed out.
Hi,
the autodiscovery feature creates the corresponding entities in my home assistant instance.
However, it does not create a device that accumulates all the entities.
Is this the behaviour everyone experiences?
Is there a way to remove specific entities from home assistant?
I’m trying to remove an old match, but the home assistant entity has the remove button grayed out.
This should not be DoubleTake specific.
Do a search for that topic in general, as this is a common issue in home assistant and multiple option to remove these entities exist. Which one particularly applies to your case, you will need to check.
Also, check that the entity message is not retained in your MQTT server.
Can you maybe put your code in a code block? Tried to make it work but I could not find the way to fix the errors. Do you also have an action part?
I’m sure I’m just using the wrong key words when I try to search for this:
but how can I reduce the timing on this?
Right now I have about 40 pages of images on Double Take for the last couple of hours. Most are just a few seconds apart.
For certain cameras, like the living room, can I make it so that Frigate only sends anything to DoubleTake every 10 or 30 or 60 seconds instead of constantly? But only for certain cameras, the ones outdoors I would like to run more frequently.
Hello, just started with frigate and face detection and saw double take is no longer developed (no updates for 2 years), deep stack has no updates, facebox has no dev account anymore and gets removed.
Compreface would be nice but is not running on my NUC with Intel Core because no AVX support. Or is there anywhere an add-on for HA that runs with CPU or Coral TPU instead of AVX?
Or do you know a different face detection add-on for home assistant that works good and is still supported?