Thanks for the interesting additional info. Fortunately 120sec is fine for my use case. I suppose since I'm only delaying BI service startup (AFAIK anyways, lol), maybe there is room to determine how long it has to wait... and maybe tune that 120 down a bit.
Blue Iris and CodeProject.AI ALPR
- Thread starter MikeLud1
- Start date
From docker reading, if windows is installed on exsi vm it will not work.
Sent from my SM-S906U using Tapatalk
Hey @MikeLud1 even after adding "alpr,Plates" to my custom models the AI analysis still shows red X's.
Am I missing something in the settings?
And while I have your attention, I'm still running CP.AI V2.0.6, mostly due to how long it has taken to install/upgrade CP.AI over my slow DSL connection.
Am I really missing anything by not upgrading to V2.0.7? If 2.0.7 is as described on the CP.AI website "2.0.7: Improved logging, both file based and in the dashboard, module installer/uninstaller bug fixes " I'd rather wait for a release that has functionality I need rather than risk breaking what seems to be working OK for me.
You need to add
,* to the end of the To Confirm list, so it should be person,vehicle,NightPlate,DayPlate,*That did give me a Green check mark for a plate, but appears to have a bit of a strange side-effect. I have a camera in a store room off of my office that I added alpr to so I can test CP.AI, including alpr, by walking into the room holding an old license plate.
With
,* added to the end of To Confirm, if any of the custom models respond with Nothing found for the alert, "Nothing found" gets added to the alert in the BI Console.Walking into view of the camera without holding a plate, the alert gets tagged:
Sticking the plate into the camera's field of view without ipcam-general recognizing a person, the alert gets tagged:
And did you have an opinion on upgrading from 2.0.6 to 2.0.7 if 2.0.6 seems to be working ok?
@MikeLud1, little bit confused here. why do need ,* on the confirm. I also noticed under customer models alpr and plates, what are they for? I was running it without them and I got under alerts the plate and the plate number (By the way only on the BlueIris not on the mobile unfortunately) I cloned the camera just for plate recognition, not sure if it makes sense.
Thanks
Thanks
Attachments
Do you happen to have the YAML for creating the entity that MQTT will feed with the data at each alert? Or do you mind sharing how you might feed HA? Appreciate any help !
FWIW, I removed
,* and it's all been working very well for a few days straight. It deals with parked cars as expected etc.One suggestion for folks wanting their LPR cam to perform multi-role, use a clone for that. I previously had a similar setup, using combined and alpr on one camera, with person, Nightplate, etc to confirm. Several days ago I changed that so the main cam only has alpr and only confirms day/night plates. I created a clone of the camera and set it up with just combined with person. Since doing that, I no longer have random issues with make times spiking to like 2sec... even when there is a lot of action the separate clone with separate motion trigger and AI confirm seems to be working more efficiently for me. I have no clue why though, lol... just my observations.
@truglo, thank you for your suggestion, it makes sense. Quick question for you since I'm still experimenting, what is the difference between license plate and aplr, I have attached the screen few post above this one.
Thanks for those posts, they are very useful.
Sent from my SM-S906U using Tapatalk
Thanks for those posts, they are very useful.
Sent from my SM-S906U using Tapatalk
Thanks. I failed through this and came up with the below under the MQTT of my YAML. I can update from BI, however I am not sure what the variable is to feed it with. I tried &PLATE with no success. Does anyone know what the right thing is to send the plate info from BI?
sensor:
- name: LicensePlateALPR
state_topic: blue_iris/LicensePlateTpole/state
Vettester
Getting comfortable
- Feb 5, 2017
- 822
- 823
Payload: { "plate":"&PLATE", "AlertImgPath":"&ALERT_PATH", "Alert_AI":"&MEMO", "Date":"%Y-%m-%d %H:%M:%S","Camera":"&CAM"}
Out of curiosity (and lack of creativity), I have to ask you guys... what are you doing with plates in HA?
Only thing I conjured up in a few minutes of thought is maybe a voice alert when an expected visitor arrives... but I suspect you all have much cooler applications.
Only thing I conjured up in a few minutes of thought is maybe a voice alert when an expected visitor arrives... but I suspect you all have much cooler applications.
funny, that’s exactly what I was doing when I used the online version ALPR. I registered lots of plates and tied them to names of companies and my outdoor speakers would greet them by name or by company. I also would automate lights, temp, etc if it was our cars pulling into the garage. I hated That the in-line ALPR stored all your images so I ditched it and cleared all my configs. If I knew I would eventually be able to run it locally I would have kept it all instead of re-failing through it.
Also @Vettester, I should have said that I do basically enter what you shared but it isn’t sending the data even though it’s in the alert data.
Here's am interesting one, reads the grill of the vehicle as a License Plate while not reading the actual plate.
Vettester
Getting comfortable
- Feb 5, 2017
- 822
- 823
Vettester
Getting comfortable
- Feb 5, 2017
- 822
- 823
Something must be misconfigured somewhere. Your MQTT message should look like this:
I'll try to take more stabs at it this weekend. One thought I had is maybe my alert sends before a plate number is identified. I'll have to add DayPlate,NightPlate to the AI requirements of the trigger.
Thanks for the responses !
Thanks for the responses !
Vettester
Getting comfortable
- Feb 5, 2017
- 822
- 823
I mean... kinda does look like a plate I guess? It would be even better if it read LOLOLOL though.
Check this one out from today...
This was from a clone of my LPR running ipcam-combined. The master camera did get the plate as well, but had an L instead of a 4 (which is very common BTW). That's why the return times are a bit up there I think... was processing OCR at the same time.