Face and person detection with Deepstack - local and free!

Hello, thanks for using DeepStack. I would like to know, did you test the new DeepStack on the Core i7 you used before?

Itā€™s really interesting to see the tests you did. Thanks a lot for the feedback. We will fix the bug causing the server to crash.
We would also investigate the issues with the RAM.

DeepStack will be eventually open source.
A new version will be up in a few days, with a lot more features and improvements.

thanks for the quick response, the issue is not with the RAM,
By saying it takes ~1.2GB I meant just 1.2GB, I was expecting for much more, especially after using DLib over the last year.

The issue was when I increased both CPU cores and RAM responses were slower (~1.5 seconds),
My PC is running i7 8700T if thatā€™s assists.

If you need any logs just let me know

1 Like

Hi, is it possible to launch this command when HASS restarts (automatically), may you please tell me how?

Sometimes I see the service is stopped and I donā€™t want to reboot the host, since I do restart HASS once a day or two, I was hoping to make this service restart too (automatically)

EDIT when I say service stopped, I mean I read UNKNOWN in frontend

You can install portainer to manage it, one of the flags you can set for the docker is restart, if you are using the api of docker or portainer you can use the restart endpoint.

There is an issue I explained few posts above that it crashes due to an unportected scope that fails for time to time (depends on the number of calls to recognize),
I got to the code that runs in the docker (also explained where it happens) and wrapped it with try catch, it solves that issue.

Hope that in the next version it will be solvedā€¦

Just installed and ā€¦ (but 5002 is deepstack port ā€¦)

Failure

driver failed programming external connectivity on endpoint deepstack (8855aad5af2xxxxxxxxxxxxxxxxx): Bind for 0.0.0.0:5002 failed: port is already allocated

EDIT actually I have so many deepstack but stopped, not sure why

[thirsty_pascal](http://192.168.1.1:9000/#/containers/827906aefa7b9dc2d00b)
[confident_haslett](http://192.168.1.1:9000/#/containers/06fdcc6d 809ef9097acfe958215bf7c)
[deepstack](http://192.168.1.1:9000/#/containers/63708bca024dc 56a35bb5afe08b5224)

This one is running, and I restarted it now. Any chance to make this automatic?

[musing_feynman](http://192.168.1.1:9000/#/containers/6d7267a1202 b425fa7525b2cfba6ed2)

the port you are trying to allocate already being used, try another port,
Iā€™m using simple mapping 0.0.0.0:5000ā€“>5000.

Iā€™m not at home right now, so I donā€™t remember how exactly the tabs named, probably restart (2nd/3rd tab from the end), you can set in the sudo docker run command the flag:
ā€“restart always

Following your log messages, Iā€™m not sure it will assist, can you extract the logs from within the docker (once the container is running you have an option to go in Portainer to extract the logs of the container).

Thanks for the extra info. Would investigate this.

How come the cpu-load is so high when no images are being processed?

To activate the licens i am trying to navigate to the url ā€œhttp://10.0.0.101:5000/adminā€ but keep getting 404 page not found.

The log shows

Hi,
I updated deepstack to the latest version and now I am getting this error when trying to start the docker container:

deepstack        | Mar 17 12:07:31 5837a95b3c20 cron[25]: (CRON) INFO (pidfile fd = 3)
deepstack        | Mar 17 12:07:31 5837a95b3c20 cron[25]: (CRON) INFO (Running @reboot jobs)
deepstack        | panic: runtime error: slice bounds out of range
deepstack        |
deepstack        | goroutine 1 [running]:
deepstack        | main.decrypt(0xc0001cdec8, 0x10, 0x20, 0x0, 0x0, 0x20, 0x0, 0x10, 0xb8a000)
deepstack        |      /home/john/Documents/AI Commons/DeepStack/DeepStack February/DeepStack/server/server.go:320 +0x39c
deepstack        | main.main()
deepstack        |      /home/john/Documents/AI Commons/DeepStack/DeepStack February/DeepStack/server/server.go:2083 +0x118a
deepstack        | panic: runtime error: slice bounds out of range

Any idea what could cause it? I am running it on a Intel nuc.

Many thanks

1 Like

I encountered the same issue after downloading the latest image (also tried cpu-3.0).

@johnolafenwa, should we continue reporting those issues here or in the new forum of deepstack.cc?

thanks

@johnolafenwa I reduced 1.5 seconds (out of 3.9 seconds in average) from each call by commenting in the intelligence.py the call for load_faces in each call, I saw itā€™s being called when the code is starting to run and everytime register is being called, sorry for getting into your code, but Iā€™m trying to assist :slight_smile:

EDIT: Find out that I missed the place where it should load the faces, added it and now the improvment is not that big, just 0.2 secondsā€¦

Hello, Please visit the localhost:5000 instead. Also, for future issues, please post to https://forum.deepstack.cc

Thanks

Hello, sorry for the issue, it was a bug that affected those who activated the previous beta release.
A new update has been issued to fix this. Please pull the latest version.
For future issues, please post on https://forum.deepstack.cc

Thanks for using deepstack

Hi, nice work on the tests. A new update has been pushed to fix the issue. Just pull the latest image or the cpu-3.1 tag and it will run fine.

You should post future issues on https://forum.deepstack.cc

Also, DeepStack will be open sourced in a few weeks from now. Future versions of it will be community driven.

We are planning a new version that will be almost twice as fast as the current version. That will be out in about two weeks maximum. Thanks for trying out deepstack.

1 Like

How can I access localhost:5000 when I am running headless?

thanks, it works great!

Waiting for the next version :slight_smile:

2 Likes

Just to make sure

Mode:
high - Low accuracy?
medium - Normal?
low - High accuracy?

also how to change confidence level manually?

you can use a ssh port forward/tunnel from your SSH client.
E,g, if using putty, in the config/SSH/tunnels, set a source port, normally the same as the one you want to access but doesnā€™t have to be, so 5000, and then the destination would be 127.0.0.1:5000. Press ā€˜addā€™, press ā€˜okā€™ and then assuming all the ports are free itā€™ll forward all requests from your client on the ā€˜sourceā€™ port to the server using the destination configuration

1 Like

Thanks, will try that!