exadel-inc / CompreFace

Leading free and open-source face recognition system
https://exadel.com/accelerator-showcase/compreface/
Apache License 2.0
5.69k stars 775 forks source link

Any chance for 1.2 as home assistant Addon? #1218

Open Mikescotland opened 10 months ago

Mikescotland commented 10 months ago

Any chance for 1.2 as home assistant Addon? Currently on 1.1 since last year....

desheh commented 10 months ago

+1.

There's the one in this repo but it's stuck at 1.1 and seems like the owner of that repo isn't maintaining it.

Would probably be best if you provided it yourselves, assuming that's something you can take on naturally.

Mikescotland commented 10 months ago

That's fine. I forced myself and finally installed in a docker!

backslashV commented 9 months ago

I am not sure if this has done it, but I simply updated config.json and Dockerfile from jakowenko's repo to point to v1.2.0, and it seems it's working. The main reason I wanted to move away from v1.1.0 was the huge amount of memory the addon was using, apparently due to memory leaks in that version. However, with v1.2.0, I am seeing an even higher memory usage at a constant 4GB. Before, it was around 2.7GB. @pospielov do you mind sharing some insight in this regard? Is this amount of memory usage normal, or is the way I am creating v1.2.0 of the addon wrong?

wills106 commented 9 months ago

There was this advise when 1.1.0 came out. I'm using it on a GTX1650 with only 4GB VRAM https://github.com/exadel-inc/CompreFace/issues/1034#issuecomment-1481932849

I'm running this on Unraid using docker compose with the following:

ADMIN_VERSION=1.1.0
API_VERSION=1.1.0
FE_VERSION=1.1.0
CORE_VERSION=1.0.0-arcface-r100-gpu
POSTGRES_VERSION=1.1.0

I haven't tried updating any of the individual modules to 1.2.0 yet. Also it looks like that addon uses a single container version. So I'm not sure how you would split it like compose.

Edit: As you only have the single tag to choose from you might be best using: 1.0.0-mobilenet-gpu or 1.0.0-arcface-r100-gpu If your system has a GPU?

backslashV commented 9 months ago

There was this advise when 1.1.0 came out. I'm using it on a GTX1650 with only 4GB VRAM #1034 (comment)

I'm running this on Unraid using docker compose with the following:

ADMIN_VERSION=1.1.0
API_VERSION=1.1.0
FE_VERSION=1.1.0
CORE_VERSION=1.0.0-arcface-r100-gpu
POSTGRES_VERSION=1.1.0

I haven't tried updating any of the individual modules to 1.2.0 yet. Also it looks like that addon uses a single container version. So I'm not sure how you would split it like compose.

Edit: As you only have the single tag to choose from you might be best using: 1.0.0-mobilenet-gpu or 1.0.0-arcface-r100-gpu If your system has a GPU?

Thanks. There is no .env file for the addon. I am not sure where to put those configs. I am running Home Assistant on an Intel NUC with an integrated GPU. I also have a Google Coral attached, but I don't think CompreFace supports it.