Is your deepstack running on Docker? You need to use the :noavx tag when pulling the image or you'll ge the above error
[tool] [tutorial] Free AI Person Detection for Blue Iris
- Thread starter GentlePumpkin
- Start date
i have a question about the url,
if an motion is detected i can add a url which is executed, can i put a variable with the name of the actual image?
if an motion is detected i can add a url which is executed, can i put a variable with the name of the actual image?
freefaller85
n3wb
Yes. I ran both to no avail. Get the same error with both the :latest and :noavx. I also tried running both in Portainer in Home Assistant on a different PC with the same results.
Last edited:
This may be nothing, but in windows docker,it runs in 2 modes linux platform or windows platform, I'm running my deepstack docker in linux platform. I've never tried it running in windows platform, don't know if it would run perfectly or run with errors in windows platform.
if you are using docker desktop gui rick click you should see something. I use the command line i specify volume mount point, which I can't do in docker desktop.
i also mapped the port to 8080 for AI Tool to use. e:\dockerData is folder i precreated before running docker for deepquestAI to store whatever data it wants there permanently.
i got this working only couple weeks ago, though the image I downloaded 5 weeks ago.
docker run --platform linux -dti --name "deepquestAI" --restart always --net="bridge" -v e:/dockerData:/datastore -e MODE=Medium -e VISION-SCENE=True -e VISION-FACE=True -e VISION-DETECTION=True -p 8080:5000 deepquestai/deepstack:latest
pmcross
Pulling my weight
It should be able to be. AI Tool just watches a folder for snapshots that were generated by a motion event. I’m not sure if the NVR supports dropping images into a folder on a Windows machine but, at a minimum, you could have each camera do this. You could then receive alerts from AI Tool via Telegram, or you could trigger the camera using the Url to the NVR.
Sent from my iPhone using Tapatalk
I've had this successfully working using a docker image on windows and setting it to port 80, but any time I try to use a port other than 80, it fails to get a response from deep stack. Seems like everyone else is able to use a non port 80, so wondering if anyone has a clue of what I'm missing... tried looking at firewall ports, etc...
Thanks for any thoughts.
Code:
ERROR: Processing the following image 'D:\BlueIris\AiInput\ai_test.20200605_083635650.jpg' failed. Can't reach DeepQuestAI Server at http://192.168.xxx.xxx:83/v1/vision/detectionThanks for any thoughts.
sunnyhaibin
n3wb
- Jul 12, 2017
- 12
- 0
Refer to post #338 ([tool] [tutorial] Free AI Person Detection for Blue Iris) - I'm using 1.63 cause I could not get 1.65 to run on all of my machines, but one of the machines had the exact issue like yours, and installing the mentioned Visual C++ resolved the issue after a couple of restarts of AI Tool and Deepstack server.
I'm not sure how that would be related to the port issue. Deep Stack is running fine and can be accessed via the port in a web browser.. I've also installed a docker image under home assistant and get the same result from AI Tool...seems to be an issue with AI Tool not connecting to the right port... and why would it work on port 80 just fine?
Anyway, I'll give it a shot and see if it changes anything...
thanks for the idea!Try starting docker with Network in Bridge mode. In Bridge mode the port 80 will be listened not just at your 192.168.xxx.xxx machine ip, but also 127.0.0.1 if you are running AITool on the same
machine, if your issue is W10 firewall, maybe the firewall wont block if your Deepstack url in AItool is changed to 127.0.0.1:80 or another port # of your choice.
I think firewalls are less restrictive to listening ports on loopback (127.0.0.1) interface
you can also try port 8080 Some firewalls treat port 8080 similar to 80.
Yeah, nothing obvious there either unfortunately. Guess I should be content I have it working on port 80 for now.
Thanks!
1 more trick you can try . click Windows key then type -> Windows Defender with Advanced Security
then click on->Inbound Rules
you'll see somewhere in inbound rules 2 com.docker.backend
1 is for tcp
1 is for udp
double click on the row for tcp
General tab-> change action from "block the connection" to "allow the connection"
Protocols and Ports tab-> change Local port from 80 to All ports
click ok
I may have changed mine a long time ago.
i don't recall if I had to reboot for it to take affect.
then click on->Inbound Rules
you'll see somewhere in inbound rules 2 com.docker.backend
1 is for tcp
1 is for udp
double click on the row for tcp
General tab-> change action from "block the connection" to "allow the connection"
Protocols and Ports tab-> change Local port from 80 to All ports
click ok
I may have changed mine a long time ago.
i don't recall if I had to reboot for it to take affect.
Mine's already set that way... it's probably another setting like that somewhere I'll have to dig for when I have more time... but I've tried docker on the same machine, Virtualbox on another machine, same deal....80 works. Thanks for the suggestions!
Got my Google Coral USB Accelerator yesterday. I currently have it parsing the rtsp streams. Getting detection times of <30 ms.
Is that still via deepstack? I thought the coral models were generally TF lite and far less accurate
TF lite. Have had no issues yet. I am doing a fait bit of post processing though (e.g. non maximum suppresion, object has to be found in 3 consecutive frames, ...)
Also, it is not that big of a deal. I can process nearly every single frame now, so i can compensate for accuracy if needed.
That's the natural next step - just process every frame, and you don't need to run/tune a separate motion detector
