Change of Parameter Range

[demenzio]

The Feature

In an earlier version, there was no limit for the value entered in the “Analog/Digital Transition Start” item.

Now I have tweaked the configuration again. But now I have the problem, that I can no longer enter the correct value for me, because the allowed limit is between 6.0-9.9.

For me, however, it is now the case that the first digit before the decimal point starts to rotate at 0.50 and finishes rotating at 0.59/0.60.

That's why from 0.6+ I get one m^3 too much and later a negative value.

[SybexX]

You can edit/remove this limitation by editing the edit_config.html with a text editor.
Then simply set the two values ​​to your desired ones.

[caco3]

Or even easier, edit the config file directly.

I understand that the limitation is annoying for you, but it is there as a guidance for less experienced users.
How ever I could accept to change the min value from 6.0. to 5.0.

[demenzio]

Well that would have been a way faster way. Next time I will do that. :)
I changed the following files as shown above:

• edit_config.html
• edit_config_template.html

It is not my main problem. I just thought it would be the main cause of the false post-processing.

I waited for it to go up to 0.5+ and that happened to day, but it is still the same issue:

As you can see, it calculates a wrong value in post-processing. It needs to be corrected to 299, because the first digit is already fully turned (turning cycle between 0.5-0.6).

An now I don't know what setting to set.

Found someone with a similar issue

Config:

[TakeImage]
;RawImagesLocation = /log/source
WaitBeforeTakingPicture = 5
;RawImagesRetention = 15
Demo = false
Brightness = 0
Contrast = 0
Saturation = 0
Sharpness = 0
LEDIntensity = 22
ImageQuality = 12
ImageSize = VGA
Zoom = false
ZoomMode = 0
ZoomOffsetX = 0
ZoomOffsetY = 0
Grayscale = false
Negative = false
Aec2 = false
AutoExposureLevel = 0
FixedExposure = false

[Alignment]
InitialRotate = 184.4
InitialMirror = false
SearchFieldX = 20
SearchFieldY = 20
AlignmentAlgo = default
FlipImageSize = false
/config/ref0.jpg 167 247
/config/ref1.jpg 303 92

[Digits]
Model = /config/dig-cont_0620_s3_q.tflite
CNNGoodThreshold = 0.5
;ROIImagesLocation = /log/digit
;ROIImagesRetention = 3
main.dig1 211 157 20 36 false
main.dig2 241 157 20 36 false
main.dig3 271 157 20 36 false
main.dig4 301 157 20 36 0
main.dig5 331 157 20 36 0

[Analog]
Model = /config/ana-cont_1208_s2_q.tflite
;ROIImagesLocation = /log/analog
;ROIImagesRetention = 3
main.ana1 399 217 71 71 false
main.ana2 363 300 71 71 false
main.ana3 279 334 71 71 false
main.ana4 190 296 71 71 false

[PostProcessing]
main.DecimalShift = 0
main.AnalogDigitalTransitionStart = 5
PreValueUse = true
PreValueAgeStartup = 720
main.AllowNegativeRates = false
main.MaxRateValue = 0.3
;main.MaxRateType = AbsoluteChange
main.ExtendedResolution = false
main.IgnoreLeadingNaN = true
ErrorMessage = true
CheckDigitIncreaseConsistency = false

[SybexX]

Once you have edited the edit-config.html, you don't need to change anything in the config.ini
edit_config_template.html is only needed to build the edit_config.html and is not used/required by the ESP32-Cam

[Slider0007]

It is not sufficient to change limits in HTML part only because it's limited in firmware as well.

AI-on-the-edge-device/code/components/jomjol_flowcontroll/ClassFlowCNNGeneral.cpp

Line 251 in 8481cc4

if ((eval_predecessors>=6 && (numeral_preceder>analogDigitalTransitionStart || numeral_preceder<=0.2) && roundedUp)) {

I did some tests with this part of logic in this branch:
https://github.com/jomjol/AI-on-the-edge-device/commits/analog-digit-early-digit-test/
I lowered the limit to 3 for testing purpose. Additionally I added some logic to potentially handle such a situation where last digit zero crosses to early: #3028

@haverland: This could be evaluated further and possibly added to firmware. While the lenghty discussion with rainman I added some test cases in the branch mentioned above. Maybe you can have a look to it.

[haverland]

It is not sufficient to change limits in HTML part only because it's limited in firmware as well.

AI-on-the-edge-device/code/components/jomjol_flowcontroll/ClassFlowCNNGeneral.cpp

Line 251 in 8481cc4

if ((eval_predecessors>=6 && (numeral_preceder>analogDigitalTransitionStart || numeral_preceder<=0.2) && roundedUp)) {

I did some tests with this part of logic in this branch: https://github.com/jomjol/AI-on-the-edge-device/commits/analog-digit-early-digit-test/ I lowered the limit to 3 for testing purpose. Additionally I added some logic to potentially handle such a situation where last digit zero crosses to early: #3028

@haverland: This could be evaluated further and possibly added to firmware. While the lenghty discussion with rainman I added some test cases in the branch mentioned above. Maybe you can have a look to it.

Looks good. All tests are fine. I found no uncommented test cases. So I think it should be ok.

But I don't understand what it means when analogDigitalTransitionStart can take any value between 0 and 10.

[SybexX]

https://jomjol.github.io/AI-on-the-edge-device-docs/Watermeter-specific-analog---digital-transition/

[Slider0007]

@haverland

But I don't understand what it means when analogDigitalTransitionStart can take any value between 0 and 10.

To be on the safer side I'd suggest to release only from 3-10. This is how the branch is prepared. I assume most meter should be covered with this range.

BTW: Unfortunately the parameter name is a bit misleading, because actually it is and was already programmed differently (Not the start of transition is relevant (like the name suggests), but when last digit reaches uncertainty region > x.8). Therefore in my opinion at least the documentation should be adpated as well.

[haverland]

Any suggestion for another name. Technically its correct, because we need to know the start of the transition phase of the digit.

[Slider0007]

Any suggestion for another name.

That's a good question :-) Maybe more generic "Analog Digit Sync Value".

Technically its correct, because we need to know the start of the transition phase of the digit.

This is what I also thought but as far I can see it's seems not coded like this, because the respective part of code is only active whenever the digit reaches >=x.8 due to the roundedUp flag.

Code in actual develop:

AI-on-the-edge-device/code/components/jomjol_flowcontroll/ClassFlowCNNGeneral.cpp

Line 251 in 32282ec

if ((eval_predecessors>=6 && (numeral_preceder>analogDigitalTransitionStart || numeral_preceder<=0.2) && roundedUp)) {

[haverland]

"Analog Digit Sync Value" is good. Alternatively "Analog Digit Shift" ?

Yes the code should not work, but it did in the test cases. I will have a deeper look on it next days.