Thought I would start a new thread so not to plug up the bot thread. I am getting these results running these settings on my demo.
BI6 new built in AI beta 6.0.1.0
- Thread starter Tinman
- Start date
Thought I would try it quick on my main BI, after many attempts and restarts, I never could get it to work with my GPU Nvidia RTX A2000, it just would always return nothing found. I then just tried "NO GPU" and I did manage to get it to detect, but terrible time.
NO GPU:
Back to 6.0.0.1 and using GPU:
NO GPU:
Back to 6.0.0.1 and using GPU:
MikeLud1
IPCT Contributor
The issue is the JSON format from v6.0.1, the api is just objects, not alpr. This is something that Ken will need to fix because @algertc’s alpr database app only recognizes the api alpr or @algertc’s needs to look for the label plate and use this for the plate.
v6.0.1
JSON:
LPR.20251217_211126Z.253663.3-5.dat
[
{
"api":"objects",
"found":{
"success":true,
"error":null,
"message":"",
"predictions":[
{
"category":"vehicle",
"label":"truck",
"confidence":0.802895486354828,
"x_min":5,
"x_max":1335,
"y_min":1,
"y_max":787,
"plate":"XMW6099"}
,
{
"category":"vehicle",
"label":"XMW6099",
"confidence":0.964644551277161,
"x_min":1167,
"x_max":1277,
"y_min":504,
"y_max":578}
,
{
"category":"vehicle",
"label":"car",
"confidence":0.787121057510376,
"x_min":966,
"x_max":1917,
"y_min":168,
"y_max":912,
"plate":"XMW6099"}
,
{
"category":"vehicle",
"label":"XMW6099",
"confidence":0.95627498626709,
"x_min":1165,
"x_max":1275,
"y_min":502,
"y_max":580}
]
,
"count":4,
"moduleId":"",
"executionProvider":"ONNXRunTime",
"analysisRoundTripMs":1244}
}
]My ALPR
JSON:
T+300 msec [22430 msec]
[
{
"api":"objects",
"found":{
"message":"Found car,
car",
"count":2,
"predictions":[
{
"label":"car",
"confidence":0.9540823,
"y_min":190,
"x_min":410,
"y_max":923,
"x_max":1918}
,
{
"label":"car",
"confidence":0.8068814,
"y_min":1,
"x_min":2,
"y_max":770,
"x_max":1384}
]
,
"inferenceDevice":"GPU:0",
"inferenceMs":2124,
"processMs":2124,
"analysisRoundTripMs":6736,
"success":true,
"moduleName":"ObjectDetectionYOLO11Net",
"moduleId":"ObjectDetectionYOLO11Net",
"command":"detect",
"requestId":"b5fd2c86-7beb-4f53-b532-b468a36f2d4c",
"processedBy":"localhost",
"timestampUTC":"Wed,
17 Dec 2025 21:13:58 GMT"}
}
,
{
"api":"alpr",
"found":{
"message":"Found 1 license plate(s): LP4634",
"count":1,
"predictions":[
{
"confidence":0.9014773,
"is_day_plate":true,
"label":"LP4634",
"plate":"LP4634",
"x_min":563,
"y_min":474,
"x_max":706,
"y_max":580,
"top_plates":[
{
"plate":"LP74634",
"confidence":0.9014773}
]
,
"corners":[
[
589.8317,
494.9086]
,
[
687.16003,
503.5963]
,
[
681.8087,
557.60144]
,
[
584.1287,
547.7939]
]
,
"track_id":4103,
"speed_mph":18.143211,
"tracking_frames":1}
]
,
"inferenceDevice":"GPU:0",
"inferenceMs":288,
"processMs":506,
"analysisRoundTripMs":526,
"success":true,
"moduleName":"ALPRYOLO11NET",
"moduleId":"ALPRYOLO11NET",
"command":"alpr",
"requestId":"2ad0bb71-790a-48ff-b82c-458cc87f7d8c",
"processedBy":"localhost",
"timestampUTC":"Wed,
17 Dec 2025 21:13:58 GMT"}
}
]
Last edited:
Vettester
Known around here
- Feb 5, 2017
- 1,069
- 1,067
I sent an email to Ken and he found the issue. He said it will likely be fixed in the next release (6.0.1.2).
I've installed v 6.0.1.2 and can't analyse dat files anymore. I get an error:
this seems to be related to the dat file itself, as I can analyse v6.0.0.1 dat files correctly.
Furthermore, is there a way to fine-tune ALPR with parameters like MikeLud's module with the internal ALPR? I was able to achieve very good results with CPAI module only after some fine tuning.
Also, parameter &PLATE in alert actions is empty, even if the plate has been found.
thank you!
this seems to be related to the dat file itself, as I can analyse v6.0.0.1 dat files correctly.
Furthermore, is there a way to fine-tune ALPR with parameters like MikeLud's module with the internal ALPR? I was able to achieve very good results with CPAI module only after some fine tuning.
Also, parameter &PLATE in alert actions is empty, even if the plate has been found.
thank you!
Last edited:
Vettester
Known around here
- Feb 5, 2017
- 1,069
- 1,067
6.0.1.3 was just released and the JSON format issues are resolved.
nstig8
n3wb
- Jan 21, 2015
- 24
- 11
Installed 6.0.1.2 this morning and my No GPU inference times went to around 350ms from about 200ms on the same machine with CodeProject.AI running in a docker container. The Blue Iris default model is yolov8 and my CPAI is using yolov5, so that may be part (or all?) of the difference. No obvious insight on if it works better or not, they both seem to identify me and my dog equally well. Just giving my experience since I can't find many posts talking about this yet.
EDIT: Also, the very first inferences after a restart are multiple seconds, I guess while the model loads into cache or something. Don't use those numbers for comparison. You have to get back up and walk past the camera more than once to get real data.
EDIT: Also, the very first inferences after a restart are multiple seconds, I guess while the model loads into cache or something. Don't use those numbers for comparison. You have to get back up and walk past the camera more than once to get real data.
Last edited:
Vettester
Known around here
- Feb 5, 2017
- 1,069
- 1,067
I was having the same issue and Ken told me that you can't open a dat created by 6.0.1.2 in anything earlier because the format has changed.
Last edited: