Blue Iris and DeepStack ALPR

Are you using DeepStackALP, if so how is it working for you?
At the moment I have not try this DeepstackALP model but want too.. I was having very high Response from Deepstack because of my own not knowing issues lol, like 2000 ms.. or so or even 3000ms response delays..

just yesterday I started reading thru some of this.. And started to tune. What I learned LOL I had too many tags on regular and custom models so It was throwing delays now that I got that under control.. I will probably try this soon..


What I am thinking is maybe someone with enough Programming experience .. Can somehow Make a software where they can pick up the data your model collects and than run thru a Plate Serial format.. Where than it can detect What state the plate belongs to from the Number series alone...

@DLONG2 Introduced me to this series sometime ago... This maybe helpful some-point in the future for this project..

 
I just posted V2.2 in the first post.

Thanks for your hard work on this! I am implementing it as we speak. I have two LPR cameras running in Australia, if I can contribute/help in any way please let me know

Cheers
 
I still can't figure out what my problem is. (v2.1 and v2.2) Don't get any alerts fra LPR camera, but the plates are very clear in the images. Don't get the square around the plates.
 
  • Wow
Reactions: sebastiantombs
I still can't figure out what my problem is. (v2.1 and v2.2) Don't get any alerts fra LPR camera, but the plates are very clear in the images. Don't get the square around the plates.
Are you getting cancel alerts for the LPR camera, if so post a screenshot like the below screenshot. It could be the license plates in Denmark are wider than the ones in the US which the DeepStack models are trained from.
1642262649676.png
 
Last edited:
  • Like
Reactions: sebastiantombs
I still can't figure out what my problem is. (v2.1 and v2.2) Don't get any alerts fra LPR camera, but the plates are very clear in the images. Don't get the square around the plates.
You can also try the attached licence-plate.pt, this model was trained with wide plates.

1642263380565.png
 

Attachments

It didn't change anything. still no recognition.

one of my other cameras recognizes cars etc with deepstack. So it is running.
 
DS is very case sensitive. Make sure the model names and object names are exactly the same as the name of the model file and the objects it contains.
 
The learning is done when making the "pt" file, then DeepStack tries to identify the object using what it learned from the created "pt" file

Maybe it is more accurate in finding the plates with the other pt file.

is there a way to make it learn to recognize plates at night?