Hello,
unfortunatly I can´t find any solution for my project. Maybe someone of you could help me.
I want to build a dog house in the my garden. As our dog isn´t there all the time, I want to avoid that other animals use it. But our doesn´t wear a collar in garden.
So, does anybody have experiences with face recognition for dogs e. g. with a ESP32 cam module?
Thanks
Michael
The same basic approaches used for face recognition in humans could probably be made to work with dogs.
I haven't tried the above tutorial, because from the comments, I have strong doubts that the implementation and/or method is reliable.
It might be possible to teach your dog to push a "button". Dogs can be trained to do a lot of things but not all dogs all things.
Maybe extend that to a finger print/paw print reader?
I failed programming the esp32 cam module using tensor flow, may be a problem with my IDE, i don't know. I have not tried the esp-eye, I want to do it.
Yes, thank you for the answer. That would be possible and the easiest way.
But as my former plans are to make access to the house possible. As there is only a button somebody else could use that to get access.
I never tried that. Could this really be possible? Sometimes the paws are really wet and full of mud.
Would be nice to get a feedback!
Unsure if the suggestion is helpful, but microchipping dogs with RFID is common in some parts of the world.
If your dog is, by chance, microchipped, you can get relatively inexpensive modules for reading the ID, which have UART output.
Hello Karl-Marx
A light barrier at the height of the dog's shoulders would be my idea.
Thank You. Nearly the same than the push-button. But it would not keep away thieves.
GREAT... My dog is chipped! Very nice.
Do you have more information about that?
I'm not particularly knowledgeable about it all, but I'll pass on what I remember.
From memory there are a few different brands / standards of chips used for implant. You might be able to find out which is particularly common in your region. If you have any paperwork connected to your dog (registration, veterinary info, etc), there may be some mention of the style of chip.
I wouldn't want to link specific products, but I did once buy a module through one of the "usual suspect" Chinese retailers. Searching for something like "Animal Tag Reader Module" would probably be a good start point. Pay attention to the compatible frequency / chip type, as I don't remember the modules being particularly universal. You might have to try all sort of strange search queries to hopefully come across exactly what you're looking for.
I believe a lot of the modules come with the required loop-antenna, but some don't (?). If you have some way of measuring inductance (LCR meter?), there's no reason you can't make your own wire loop at home. I'm not sure how sensitive the technology is; you might be able to use an online calculator (or.. math.., shudder) to estimate the dimensions of a suitable coil.
From memory, the read distance is not great, but not terrible. If the dog brushes past the coil, I imagine it would be close enough to get a reading. Even better would be the coil mounted around the door, with the dog stepping through, however the pre-wound coils included with modules will likely not be large enough for this.
EDIT: dog stepping through coil may not be ideal. See Post #20
A valid point, but is your aim to block other animals, or humans also? While the RFID idea is definitely interesting, it pays to remember that "a lock does no more than keep an honest man honest".
Don't know. Might be worth a try. I didn't think about the mud thing.
RFID sounds most practical. Range is pretty much whatever you like in my experience but distance costs.
Dog face recognition would be a fun project but I suspect that you would need more processing power to take the picture and analyze it. Possibly something in the house or maybe cloud based.
Judging by how often my phone fails to recognize me, if you go this route, you may need to teach your dog to enter a pin to unlock the doors 
For my own understanding, how would you go about extending the range of the system, if cost were no object?
Thank you for your efforts. I´ll go to my vet first. Maybe he has some files (frequencies) for me. I don´t have any. 
Next I´ll try to find a tag reader incl. loop-antenna.
Next step will be constructing a doog house with heating.
Looks like a really nice project for a bad-weather-season.
It's been a while and i can't find the code, but I could read my dog's chip with this adafruit shield. I'm sure other would work too...
https://www.adafruit.com/product/789
The problem I think might be that from a security standpoint. Anyone who comes close to your dog can scan and copy the RD-ID Chip. Mybe only limit access to the dog-house and not allow entrance to the human house?
The second thing is range. I don't know how far you can read the chip. But I remember I needed to scan my dog from very close range...
After quite a bit of fiddling today, I did get the proof-of-concept working again. Convincing your test-subject to play along may be easier said than done..
From today's exercise, it seems that the orientation of the coil, relative to your dog's implant, is far more crucial. This may dictate the design of your door, so it's probably worth testing your electronics earlier rather than later.
It's interesting that this shield worked for you
Although I don't doubt your results, my concern is that the linked shield is nominally for 13.56 MHz RFID / NFC applications. I believe that many (most? all?) animal microchip implants operate at frequencies near 125KHz, although this is far from standardized.
Relevant reading:
The ISO conformant full-duplex type has the greatest international acceptance. It is common in many countries including Canada and large parts of Europe (since the late 1990s).
In the United States, however, three proprietary types of chips compete along with [ISO conformant full-duplex type] ... distribution of this type has been controversial. When 24PetWatch.com began distributing them in 2003 (and more famously Banfield Pet Hospitals in 2004) many shelter scanners couldn't read them.