[tool] [tutorial] Free AI Person Detection for Blue Iris

Village Guy · Dec 18, 2020

balucanb said:
@Tinbum / @Village Guy in the Windows version of AI-Tool you can select 1 or all the different detection models from the deepstack tab, When you run the Docker version that tab is gone so how would you tell it to run more than 1 model? Do you just run sudo docker run -e XXXX-DETECTION=True -v localstorage:/datastore -p 80:5000 \ deepquestai/deepstack
multiple times and change where I put XXXX to each model you want it to use?
Thanks.

I'm personally not familiar with running more than one version of deepstack simultaneously . That said I suspect your proposal would return an error complaining that the port is already in use. Each version would need it's own port to be defined. No sure how AITool would handle that scenario.

cjowers · Dec 18, 2020

Village Guy said:
balucanb said:
@Tinbum / @Village Guy in the Windows version of AI-Tool you can select 1 or all the different detection models from the deepstack tab, When you run the Docker version that tab is gone so how would you tell it to run more than 1 model? Do you just run sudo docker run -e XXXX-DETECTION=True -v localstorage:/datastore -p 80:5000 \ deepquestai/deepstack
multiple times and change where I put XXXX to each model you want it to use?
Thanks.

I'm personally not familiar with running more than one version of deepstack simultaneously . That said I suspect your proposal would return an error complaining that the port is already in use. Each version would need it's own port to be defined. No sure how AITool would handle that scenario.

I think he is speaking of the AI model (object detection / face / scene recognition), in which case you'd probably run the command again on another port, or add all into one line, using the same format (VISION-DETECTION=True, or whatever the model is you want). It should come up on the command line once it starts, listing all API's in use.
Not sure how it works with multiple models, but i could see the objects and faces being used simultaneously for BI usage.

balucanb · Dec 18, 2020

cjowers said:
I think he is speaking of the AI model (object detection / face / scene recognition), in which case you'd probably run the command again on another port, or add all into one line, using the same format (VISION-DETECTION=True, or whatever the model is you want). It should come up on the command line once it starts, listing all API's in use.
Not sure how it works with multiple models, but i could see the objects and faces being used simultaneously for BI usage.

@Village Guy , @cjowers Thanks for replying. I agree it may not be possible to run more than one version. Village guy I suspect you may be correct or more than likely I am doing something wrong- edit Pretty positive I'm doing something wrong since I have about a 2-3 weeks of exp. with docker now.

I have been working on a custom detection model using the instructions just put out and when trying to deploy it I am indeed getting errors. I have a working vision-detection model running in/on (?) Docker desktop using port 8383, I tried to run this new one on the same port and it threw a error. The working one was running when I tried. I guess you can't run 2 instances on the same port? I stopped it and tried again, same thing error, I then tried port 80, errored out. There very well could be something else going on since I am pretty clueless concerning Docker. If anyone thinks they can figure it out, I screen shot what was going on- see attachment. Thanks

Update- As I suspected the problem was the user (me) @johnolafenwa looked at my code from the attachment and I missed a keystroke there should be a dash between cpu and 2020, so it should be cpu-2020.12 Still waiting to see if I can run more than one model at the same time....

maximosm · Dec 18, 2020

I have a BI in a pc with 10 cameras and cpu working under 12% Wenn i use The AI the cpu goes 40-80 to analyse the move and after that down to 8-12%. I dont have a gpu .The question is,if i use GPU will this help the AI not to uses CPU .Is there any tip to reduce cpu?

Village Guy · Dec 18, 2020

balucanb said:
@Village Guy , @cjowers Thanks for replying. I agree it may not be possible to run more than one version. Village guy I suspect you may be correct or more than likely I am doing something wrong- edit Pretty positive I'm doing something wrong since I have about a 2-3 weeks of exp. with docker now. I have been working on a custom detection model using the instructions just put out and when trying to deploy it I am indeed getting errors. I have a working vision-detection model running in/on (?) Docker desktop using port 8383, I tried to run this new one on the same port and it threw a error. The working one was running when I tried. I guess you can't run 2 instances on the same port? I stopped it and tried again, same thing error, I then tried port 80, errored out. There very well could be something else going on since I am pretty clueless concerning Docker. If anyone thinks they can figure it out, I screen shot what was going on- see attachment. Thanks

Update- As I suspected the problem was the user (me) @johnolafenwa looked at my code from the attachment and I missed a keystroke there should be a dash between cpu and 2020, so it should be cpu-2020.12 Still waiting to see if I can run more than one model at the same time....

To the best of my knowledge you must use a different port address for each version. Try using 8384 for the second version but you will still have an issue with aitool handling more than one port. Port 80 is probably already being used by some other app.

Why can't you incorporate everything into one module?

balucanb · Dec 18, 2020

Village Guy said:
To the best of my knowledge you must use a different port address for each version. Try using 8384 for the second version but you will still have an issue with aitool handling more than one port. Port 80 is probably already being used by some other app.

Why can't you incorporate everything into one module?

You may very well be able to that...But I wouldn't know! LMAO!! I did some more searching and found where someone else was asking a similar question to mine and the response they were given was ......"you need to add a new volume mapping to map your model directory to the /modelstore/detection directory in docker, you can enable both your custom model and the vision detection in DeepStack".... Now I need to figure out how to try that, don't suppose I can just R-click someplace and create a new folder huh? Pretty sure it involves some arcane string of characters I have no idea about.

Village Guy · Dec 18, 2020

balucanb said:
You may very well be able to that...But I wouldn't know! LMAO!! I did some more searching and found where someone else was asking a similar question to mine and the response they were given was ......"you need to add a new volume mapping to map your model directory to the /modelstore/detection directory in docker, you can enable both your custom model and the vision detection in DeepStack".... Now I need to figure out how to try that, don't suppose I can just R-click someplace and create a new folder huh? Pretty sure it involves some arcane string of characters I have no idea about.

Now mapping sounds like the way forward as you will presumably only need the one port address. Once you get this all working be ready to answer a lot of questions

balucanb · Dec 18, 2020

Village Guy said:
Now mapping sounds like the way forward as you will presumably only need the one port address. Once you get this all working be ready to answer a lot of questions

That you think I am going to get that far is nice....not sure about the validity...but nice. I am now trying to figure out how exactly to do said mapping...MTF.

BossHogg · Dec 18, 2020

Hi Guys,

I'm new to this forum and IP cameras in general, but after reading a ton about how to set it up, I want to start with a system that won't overwhelm me with false notifications. The AI tool, Deepstack and Blue Iris seems like the way to go for my situation.

I'm having a problem getting the AI tool to communicate with the Deepstack server. I've tried Deepstack on a Linux VM in docker and on the Windows 10 VM that's also running BI and the AI tool. Neither one seems to work. I think the AI tool can see the images that come from BI, but there doesn't seem to be a proper connection to Deepstack. The history tab is empty.

Below I've pasted the log for the current windows setup. It looks like an error around like 126 in the traceback. I also usually get the following:

Date

Func

Detail

Level

Source

AIServer

Camera

Image

Idx

Depth

Color

ThreadID

FromFile

Filename

2020-12-18 11:32:05 AM

GetDeepStackRun

Deepstack partially running. You many need to manually kill server.exe, python.exe, redis-server.exe

Error

AITOOLS.EXE

None

DrivewaySD

None

429

1

False

AITool.[2020-12-18].log

Please let me know if anyone sees something that could lead to a solution.

Thanks.

Date

Func

Detail

Level

Source

AIServer

Camera

Image

Idx

Depth

Color

ThreadID

FromFile

Filename

2020-12-18 9:38:25 AM

DSHandleRedisProcMSG

Debug: DeepStack>> [6020] 18 Dec 09:38:25 # no config file specified, using the default config. In order to specify a config file use 'redis-server /path/to/redis.conf'

Warn

AITOOLS.EXE

None

REDIS-SERVER.EXE

68

1

4

False

AITool.[2020-12-18].log

2020-12-18 9:38:25 AM

DSHandleRedisProcMSG

Debug: DeepStack>> [6020] 18 Dec 09:38:25 # Opening port 6379: bind 10048

Error

AITOOLS.EXE

None

REDIS-SERVER.EXE

69

1

4

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> Traceback (most recent call last):

Error

AITOOLS.EXE

None

PYTHON.EXE

89

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\python\pywrap_tensorflow.py", line 58, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

90

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> from tensorflow.python.pywrap_tensorflow_internal import *

Error

AITOOLS.EXE

None

PYTHON.EXE

91

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\python\pywrap_tensorflow_internal.py", line 28, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

92

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> _pywrap_tensorflow_internal = swig_import_helper()

Error

AITOOLS.EXE

None

PYTHON.EXE

93

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\python\pywrap_tensorflow_internal.py", line 24, in swig_import_helper

Error

AITOOLS.EXE

None

PYTHON.EXE

94

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)

Error

AITOOLS.EXE

None

PYTHON.EXE

95

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "imp.py", line 242, in load_module

Error

AITOOLS.EXE

None

PYTHON.EXE

96

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "imp.py", line 342, in load_dynamic

Error

AITOOLS.EXE

None

PYTHON.EXE

97

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> Import DLL load failed: A dynamic link library (DLL) initialization routine failed.

Error

AITOOLS.EXE

None

PYTHON.EXE

98

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> During handling of the above exception, another exception occurred:

Error

AITOOLS.EXE

None

PYTHON.EXE

99

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> Traceback (most recent call last):

Error

AITOOLS.EXE

None

PYTHON.EXE

100

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "../intelligence.py", line 13, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

101

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> from sharedintelligence.commons import preprocess

Error

AITOOLS.EXE

None

PYTHON.EXE

102

1

10

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\sharedintelligence\init.py", line 5, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

103

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> from .detection3 import DetectModel3

Error

AITOOLS.EXE

None

PYTHON.EXE

104

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\sharedintelligence\detection3\init.py", line 1, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

105

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> from .process import DetectModel3

Error

AITOOLS.EXE

None

PYTHON.EXE

106

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\sharedintelligence\detection3\process.py", line 1, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

107

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> from .utils import read_pb_return_tensors,cpu_nms

Error

AITOOLS.EXE

None

PYTHON.EXE

108

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\sharedintelligence\detection3\utils.py", line 1, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

109

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> import tensorflow as tf

Error

AITOOLS.EXE

None

PYTHON.EXE

110

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\init.py", line 24, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

111

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import

Error

AITOOLS.EXE

None

PYTHON.EXE

112

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\python\init.py", line 49, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

113

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> from tensorflow.python import pywrap_tensorflow

Error

AITOOLS.EXE

None

PYTHON.EXE

114

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\python\pywrap_tensorflow.py", line 74, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

115

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> raise ImportError(msg)

Error

AITOOLS.EXE

None

PYTHON.EXE

116

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> Import Traceback (most recent call last):

Error

AITOOLS.EXE

None

PYTHON.EXE

117

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\python\pywrap_tensorflow.py", line 58, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

118

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> from tensorflow.python.pywrap_tensorflow_internal import *

Error

AITOOLS.EXE

None

PYTHON.EXE

119

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\python\pywrap_tensorflow_internal.py", line 28, in <module>

Error

AITOOLS.EXE

None

PYTHON.EXE

120

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> _pywrap_tensorflow_internal = swig_import_helper()

Error

AITOOLS.EXE

None

PYTHON.EXE

121

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "C:\DeepStack\interpreter\packages\tensorflow\python\pywrap_tensorflow_internal.py", line 24, in swig_import_helper

Error

AITOOLS.EXE

None

PYTHON.EXE

122

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> _mod = imp.load_module('_pywrap_tensorflow_internal', fp, pathname, description)

Error

AITOOLS.EXE

None

PYTHON.EXE

123

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "imp.py", line 242, in load_module

Error

AITOOLS.EXE

None

PYTHON.EXE

124

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> File "imp.py", line 342, in load_dynamic

Error

AITOOLS.EXE

None

PYTHON.EXE

125

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> Import DLL load failed: A dynamic link library (DLL) initialization routine failed.

Error

AITOOLS.EXE

None

PYTHON.EXE

126

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> Failed to load the native TensorFlow runtime.

Error

AITOOLS.EXE

None

PYTHON.EXE

127

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> See Build and install error messages | TensorFlow

Error

AITOOLS.EXE

None

PYTHON.EXE

128

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> for some common reasons and solutions. Include the entire stack trace

Error

AITOOLS.EXE

None

PYTHON.EXE

129

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:30 AM

DSHandlePythonProcERROR

DeepStack>> above this error message when asking for help.

Error

AITOOLS.EXE

None

PYTHON.EXE

130

1

17

False

AITool.[2020-12-18].log

2020-12-18 9:38:32 AM

GetDeepStackRun

Deepstack partially running. You many need to manually kill server.exe, python.exe, redis-server.exe

Error

AITOOLS.EXE

None

132

2

1

False

AITool.[2020-12-18].log

2020-12-18 9:38:32 AM

GetDeepStackRun

Deepstack partially running. You many need to manually kill server.exe, python.exe, redis-server.exe

Error

AITOOLS.EXE

None

134

4

1

False

AITool.[2020-12-18].log

2020-12-18 9:38:35 AM

Start

5 python.exe processes did not fully start in 10110ms

Error

AITOOLS.EXE

None

136

1

4

False

AITool.[2020-12-18].log

2020-12-18 9:38:35 AM

GetDeepStackRun

Deepstack partially running. You many need to manually kill server.exe, python.exe, redis-server.exe

Error

AITOOLS.EXE

None

137

1

False

AITool.[2020-12-18].log

2020-12-18 9:46:01 AM

DetectObjects

Got http status code 'Forbidden' (403) in 86ms: Forbidden

Error

AITOOLS.EXE

192.168.10.4:83

DrivewaySD

DrivewaySD.20201218_094601131.jpg

148

1

19

False

AITool.[2020-12-18].log

2020-12-18 9:46:01 AM

DetectObjects

Empty string returned from HTTP post.

Error

AITOOLS.EXE

192.168.10.4:83

DrivewaySD

DrivewaySD.20201218_094601131.jpg

149

1

19

False

AITool.[2020-12-18].log

2020-12-18 9:46:31 AM

DetectObjects

Got http status code 'Forbidden' (403) in 9ms: Forbidden

Error

AITOOLS.EXE

192.168.10.4:83

DrivewaySD

DrivewaySD.20201218_094601131.jpg

166

1

19

False

AITool.[2020-12-18].log

2020-12-18 9:46:31 AM

DetectObjects

Empty string returned from HTTP post.

Error

AITOOLS.EXE

192.168.10.4:83

DrivewaySD

DrivewaySD.20201218_094601131.jpg

167

1

19

False

AITool.[2020-12-18].log

2020-12-18 9:47:01 AM

DetectObjects

Got http status code 'Forbidden' (403) in 18ms: Forbidden

Error

AITOOLS.EXE

192.168.10.4:83

DrivewaySD

DrivewaySD.20201218_094606131.jpg

182

1

19

False

AITool.[2020-12-18].log

2020-12-18 9:47:01 AM

DetectObjects

Empty string returned from HTTP post.

Error

AITOOLS.EXE

192.168.10.4:83

DrivewaySD

DrivewaySD.20201218_094606131.jpg

183

1

19

False

AITool.[2020-12-18].log

2020-12-18 9:47:01 AM

ImageQueueLoop

... AI URL for 'DeepStack' failed '6' times. Disabling: '

http://192.168.10.4:83/v1/vision/detection

'

Error

AITOOLS.EXE

192.168.10.4:83

DrivewaySD

None

189

0

19

False

AITool.[2020-12-18].log

Village Guy · Dec 18, 2020

@BossHogg
I propose that you remove the windows version of deepstack and install docker with deepstack running in a Linux environment. After installation, test it before integrating with aitool. The docker environment will allow you to run with current versions of deepstack that are not presently available to operating within a windows environment.

BossHogg · Dec 18, 2020

Village Guy said:
@BossHogg
I propose that you remove the windows version of deepstack and install docker with deepstack running in a Linux environment. After installation, test it before integrating with aitool. The docker environment will allow you to run with current versions of deepstack that are not presently available to operating within a windows environment.

Hi Village Guy,

I previously had An Ubuntu VM running docker and the :latest version of deepstack. that didn't work either. I'll revert back to that setup and post the log.

Actually, are you proposing i install Docker in Windows instead?

Thanks for the help.

Village Guy · Dec 18, 2020

BossHogg said:
Hi Village Guy,

I previously had An Ubuntu VM running docker and the :latest version of deepstack. that didn't work either. I'll revert back to that setup and post the log.

Actually, are you proposing i install Docker in Windows instead?

Thanks for the help.

Yes

BossHogg · Dec 18, 2020

Village Guy said:
Yes

I tried to install docker for windows and got the following error:

System.InvalidOperationException:
Failed to deploy distro docker-desktop to C:\Users\Graham\AppData\Local\Docker\wsl\distro: exit code: -1
stdout: Please enable the Virtual Machine Platform Windows feature and ensure virtualization is enabled in the BIOS.

I'm running this Win10 VM on Proxmox on a Dell R710. I know the CPUs support virtualization and it's turned on in the BIOS. is there something i need to do to "pass through" the Virtualization settings from the BIOS, through Proxmox, to the Win10 VM?

Perhaps I'll post that to r/homelab.

Thanks.

balucanb · Dec 18, 2020

BossHogg said:
I tried to install docker for windows and got the following error:

System.InvalidOperationException:
Failed to deploy distro docker-desktop to C:\Users\Graham\AppData\Local\Docker\wsl\distro: exit code: -1
stdout: Please enable the Virtual Machine Platform Windows feature and ensure virtualization is enabled in the BIOS.

I'm running this Win10 VM on Proxmox on a Dell R710. I know the CPUs support virtualization and it's turned on in the BIOS. is there something i need to do to "pass through" the Virtualization settings from the BIOS, through Proxmox, to the Win10 VM?

Perhaps I'll post that to r/homelab.

Thanks.

Google WSL requirements- it tells you the prerequisites you need to include checking/enabling the virtualization. I started here- Install Windows Subsystem for Linux (WSL) on Windows 10

Mattias Fornander · Dec 18, 2020

Just a quick note that I got DeepStack up and running on my NVIDIA Jetson I had just sitting around. AI-Tool is on the BI server but all analysis is running on the little Jetson. I previously had AI-Tool run DeepStack on my BI box but it got a little crowded on the CPU since the current Windows version is not GPU accelerated.

A few thoughts:

Running Deepstack on Jetson with a fresh Jetpack microSD card
Updated all software (sudo apt update +upgrade)
Turned off desktop environment since I'm just going to access DeepStack and I can ssh into the box if I need to fix something (sudo systemctl set-default multi-user.target)
Installed latest deepstack in docker, in High mode, and asked it to restart after machine reboot:

sudo docker run --runtime nvidia --restart unless-stopped -e MODE=High -e VISION-DETECTION=True -p 80:5000 deepquestai/deepstack:jetpack-x3-beta

The DeepStack docker image defaults to using Medium as the MODE. This means that the default Jetson server is limited to processing images no larger than 320 pixels at Medium. I'm running 4K cameras but that resolution is lost and even suboptimal for DeepStack. You should resize images close to the target processing size in BI, and not ask little jetson to do that resizing before running the image recognition.

I ended up starting my docker at HIGH mode since my cameras are already pretty tuned with motion zones in BI. At that setting I'm getting about 300ms processing time per frame:

[GIN] 2020/12/18 - 20:23:15 | 200 | 324.133383ms | 192.168.1.233 | POST /v1/vision/detection
[GIN] 2020/12/18 - 20:23:15 | 200 | 280.243879ms | 192.168.1.233 | POST /v1/vision/detection
[GIN] 2020/12/18 - 20:23:17 | 200 | 287.85692ms | 192.168.1.233 | POST /v1/vision/detection
[GIN] 2020/12/18 - 20:23:18 | 200 | 288.047127ms | 192.168.1.233 | POST /v1/vision/detection
[GIN] 2020/12/18 - 20:23:19 | 200 | 293.305335ms | 192.168.1.233 | POST /v1/vision/detection
[GIN] 2020/12/18 - 20:23:20 | 200 | 281.178667ms | 192.168.1.233 | POST /v1/vision/detection
[GIN] 2020/12/18 - 20:23:21 | 200 | 274.997808ms | 192.168.1.233 | POST /v1/vision/detection
[GIN] 2020/12/18 - 20:23:22 | 200 | 283.577667ms | 192.168.1.233 | POST /v1/vision/detection
[GIN] 2020/12/18 - 20:23:32 | 200 | 269.086322ms | 192.168.1.233 | POST /v1/vision/detection

With original size 4K images I was getting close to 900ms and with MEDIUM with scaled images I was getting around 200ms.

Also note that Jetson is running a smaller size object detection model compared to the desktop gpu and cpu so the accuracy will probably be a little worse on Jetson. You can see the DeepStack settings code below for reference:
deepstack/intelligencelayer/shared/shared.py line 61-90:

"desktop_cpu": Settings(
DETECTION_HIGH=640,
DETECTION_MEDIUM=416,
DETECTION_LOW=256,
DETECTION_MODEL="yolov5m.pt",
FACE_HIGH=416,
FACE_MEDIUM=320,
FACE_LOW=256,
FACE_MODEL="face.pt",
),
"desktop_gpu": Settings(
DETECTION_HIGH=640,
DETECTION_MEDIUM=416,
DETECTION_LOW=256,
DETECTION_MODEL="yolov5m.pt",
FACE_HIGH=416,
FACE_MEDIUM=320,
FACE_LOW=256,
FACE_MODEL="face.pt",
),
"jetson": Settings(
DETECTION_HIGH=416,
DETECTION_MEDIUM=320,
DETECTION_LOW=256,
DETECTION_MODEL="yolov5s.pt",
FACE_HIGH=384,
FACE_MEDIUM=256,
FACE_LOW=192,
FACE_MODEL="face_lite.pt",
),

The current code looks like it's using busy waiting on images so the CPU usage is a little high on the Jetson when Idling (~40%) but I'm guessing that will be fixed now that it's open source:

Jetson docker image consuming a lot of resources when idling · Issue #46 · johnolafenwa/DeepStack

Awesome work! Everything is running fine but I'm surprised the CPU is running this much when it's idling. I was looking at the code, guessing the main loop is not waiting for input but ran out of t...

github.com

It would be simpler from a hardware setup to have the DeepStack running on same box as BI but I'm a little worried that the GPU h265 4K decoding of my cams are going to be crowded by DeepStack so a separate Jetson that does extra flagging looks like a good setup for now. If Jetson blows up then I just have more false detections.

Village Guy · Dec 18, 2020

BossHogg said:
I tried to install docker for windows and got the following error:

System.InvalidOperationException:
Failed to deploy distro docker-desktop to C:\Users\Graham\AppData\Local\Docker\wsl\distro: exit code: -1
stdout: Please enable the Virtual Machine Platform Windows feature and ensure virtualization is enabled in the BIOS.

I'm running this Win10 VM on Proxmox on a Dell R710. I know the CPUs support virtualization and it's turned on in the BIOS. is there something i need to do to "pass through" the Virtualization settings from the BIOS, through Proxmox, to the Win10 VM?

Perhaps I'll post that to r/homelab.

Thanks.

All bets are off if you are not running within a windows 10 Native environment. There are just too many variables.

cjowers · Dec 18, 2020

maximosm said:
I have a BI in a pc with 10 cameras and cpu working under 12% Wenn i use The AI the cpu goes 40-80 to analyse the move and after that down to 8-12%. I dont have a gpu .The question is,if i use GPU will this help the AI not to uses CPU .Is there any tip to reduce cpu?

-try deepstackai on 'low' mode
-try lower frequency of images being anaylized
-try lower resolution images from a substream camera stream
-try less cameras triggering deepstack
-try latest deepstack versions
-try a jetson nano

gpu will definately help, its way more efficient at this type of computation (but you have to run the gpu deppstack version)

Mattias Fornander · Dec 18, 2020

maximosm said:
I have a BI in a pc with 10 cameras and cpu working under 12% Wenn i use The AI the cpu goes 40-80 to analyse the move and after that down to 8-12%. I dont have a gpu .The question is,if i use GPU will this help the AI not to uses CPU .Is there any tip to reduce cpu?

Are you reducing the size of your saved images? Make BI resize them to 640x480 or even 320x240 if you you're running on default Medium MODE.

kosh42efg · Dec 18, 2020

@Mattias Fornander - I also offloaded my Deepstack to a Jetson when they released a version that finally worked on the platform! Thanks for the tip about MODE=High, I'd missed that.

Mattias Fornander · Dec 18, 2020

Mattias Fornander said:
sudo docker run --runtime nvidia --restart unless-stopped -e MODE=High -e VISION-DETECTION=True -p 80:5000 deepquestai/deepstack:jetpack-x3-beta

Apologies, don't use the Jetson beta image btw, their release image is newer:

Code:

sudo docker run --runtime nvidia --restart unless-stopped -e MODE=High -e VISION-DETECTION=True -p 80:5000 deepquestai/deepstack:jetpack

[tool] [tutorial] Free AI Person Detection for Blue Iris

Pulling my weight

Getting the hang of it

Getting the hang of it

Attachments

Young grasshopper

Pulling my weight

Getting the hang of it

Pulling my weight

Getting the hang of it

n3wb

Pulling my weight

n3wb

Pulling my weight

n3wb

Getting the hang of it

n3wb

Pulling my weight

Getting the hang of it

n3wb

n3wb

n3wb