@pfried thanks for the clear explanation

Regards

Hi @jomjol, I’ll appreciate if you can me help with this again. Frequently I get this error when capture an image:

712685 00712685 Error - RateTooHigh (2320.0000) 715005.0<p>Aligned Image: <p><img src=/image_tmp/alg.jpg></img><p>Digital Counter: <p><img src=/.....

What does this mean? How can I avoid it?

Could you give me any cue?

Regards,

Hi @jomjol, having solved the WiFi problem, it seems that the system has already been working stable for a week.
Now I would like to refine the recognition of some digits that the system cannot detect correctly. I have seen that in other cases you have asked for a sample of images of the digits to be able to train the system. I have created a .zip with all digits (arranged in folders) that the system has not recognized and that are stored in the /home/pi/water-meter/code/app/log/digital_digit/10 folder. It basically happens with the digits in Red.

I hope this information is useful you to be able to carry out the training of the system.
I would be very grateful if you could get it.

Thank you very much

Numbers.zip

Hi @jomjol I have had the service working more or less correctly for about 4 days, obtaining a sample every 30 min.

For a couple of days, when launching the request:

http://192.168.1.234:4002/wasserzaehler.html?single

, it is very frequent to get the error:

Error - Problem during HTTP-request - URL: http://192.168.1.213/capture_with_flashlight?quality=10&size=SVGA

It is very frustrating not knowing what to do or why this error appears. I restart the ESPCAM but in one or two cycles the error appears again

Could you give me a hint? How to avoid this error?

BR

@ramon2krpi3 This is how I did:

1. Download the raspi-rolling image: sudo docker pull jomjol/wasserzaehler:raspi-rolling

2. Over my old folders structure I created a new app folder and I copied my old /config & /log folders

3. I downloaded “Train_CNN_Digital-Readout_Version_6.2.0.tflite” and put it in “/home/pi/water-meter-system-complete/code/app/config/neuralnets”

4. I edited “/home/pi/water-meter-system-complete/code/app/config/config.ini” and changed the new Train_CNN tflite

[Digital_Digit]
names = D1, D2, D3, D4, D5, D6, D7, D8
modelfile = ./config/neuralnets/Train_CNN_Digital-Readout_Version_6.2.0.tflite
LogImageLocation=./log/digital_digit

5. I run the docker and it seems do it well:

pi@pi4:~ $ docker run -p 4000:3000 --mount type=bind,source=/home/pi/water-meter-system-complete/code/app/config,target=/app/config --mount type=bind,source=/home/pi/water-meter-system-complete/code/app/log,target=/app/log --memory-swap -1 jomjol/wasserzaehler:raspi-rolling
92.0
Start Init Zaehlerstand
Analog Model Disabled
Digital Model Init Done
Digital Model Init Done
Prevalue loaded from file: 00000690.0000
Wasserzaehler is serving at port 3000

6. When I ask for a reading “http://192.168.1.234:4000/wasserzaehler.html&full” the docker gives me this error:

Exception happened during processing of request from ('192.168.1.231', 59241)
Traceback (most recent call last):
File "/usr/lib/python3.7/socketserver.py", line 316, in _handle_request_noblock
self.process_request(request, client_address)
File "/usr/lib/python3.7/socketserver.py", line 347, in process_request
self.finish_request(request, client_address)
File "/usr/lib/python3.7/socketserver.py", line 360, in finish_request
self.RequestHandlerClass(request, client_address, self)
File "/usr/lib/python3.7/socketserver.py", line 720, in init
self.handle()
File "/usr/lib/python3.7/http/server.py", line 426, in handle
self.handle_one_request()
File "/usr/lib/python3.7/http/server.py", line 414, in handle_one_request
method()
File "./wasseruhr.py", line 37, in do_GET
GlobalError = wasserzaehler.CheckError()
File "/app/lib/ZaehlerstandClass.py", line 102, in CheckError
if self.readAnalogNeedle.GlobalError:
AttributeError: 'Zaehlerstand' object has no attribute 'readAnalogNeedle'

I'm lost. Do I have to modify/install anything else?
I didin't do nothing whit Tensorflow. May I do anything to manage TensoflowLite?

Regards

Hi @jomjol, could anybody summarize the step by step migration to rolling version?
Thanks!

Hi @jomjol, luckily i just found the solution. Diving over the 1,000 thread post I have found the solution. The problem is that [alignment]
initial_rotation_angle = 92 cannot have a negative angle. The configurator "http://wasserzaehler.ignorelist.com" inserts the value in negative in the "config.ini" and therefore deforms. Just edit the "config.ini" and remove the negative sign

The problem now is the digit recognition. I know I've to solve the reflection, but in the meantime I wanted to test the rest of the digits unaffected by the reflection. The first 5 should work. The third digit, 6, has a strange shape in the upper left corner and does not detect it, it assigns a 0. On the other hand, 9 sometimes detects it as 0 and other times as 6.

how can I get it to detect these digits?

Are they in a format recognizable by your trained system?
I am using the Train_CNN_Digital-Readout_Version_6.0.0.h5

Regards!

Hi @jomjol, I am implementing the project and I am encountering some difficulties. I would appreciate it if you could help me.

After generating the config.ini I cannot read any digit. When I see “… / roi.html” I get a distorted image that I show below. I have tried to choose and mix different texts in the image to form the triangle of references. For example I have used the JS2.5-04, x0,0001, 397, ... labels as references, but I can't get the image right. I was also surprised that when I launch “… / roi.html” the red boxes of ref0, ref1 and ref2 do not appear on the picture. Would be there any relationship?

Regards,

The references have been taken after leaving the holder installed and with the same flash lighting that the ESP32 LED provides.

The Docker is running on a raspberry pi4 2GB with the following structure:

sudo docker run -p 4000:3000 --mount type=bind,source=/home/pi/water-meter-system-complete/code/config,target=/config --mount type=bind,source=/home/pi/water-meter-system-complete/code/log,target=/log jomjol/wasserzaehler:raspi-latest

The counter is different from yours, with 8 digits, 5 black on white and 3 red on white.

1. Could you help me get the image right with the references?
2. Do you think there will be a problem with the type of digits that are slightly different from your project?

Regards