- Correct error in docker commmand (remove space character in bind)
- Internal update of config.ini handling (started)
- Update roi.html
- Modification of Errortext in case RateToHigh, NegativeRate (Blank after "Error", e.g. "ErrorRateToHigh" --> "Error - RateToHigh")
- Change MaxRateValue from 0.1 to 0.2 in Default Config.ini
- Update CNN-Digital to v5.0.0 (improved training data from iobroker users)
- Update CNN-Analog to v5.0.0 (improved training data from iobroker users)
- Update CNN-Digital to v4.2.0 (improved training data from iobroker users)
- Extensdion to json output: ../wasserzaehler.json
- Integration of storage of prevalue in file to reload on startup
- Correction of drawing analog counter ROIs even if they are disabled
- Raspberry Version: Remove autorestart (not working) - instead: use cron job for regular restart to handle tensorflow memory leak Setting up cron job
- Raspberry Version: Autorestart on Python Crash of wasserzaehler.py
- Separate environmental setup to dedicated Docker images (for Raspberry: raspi-opencv-tensorflow and for Synology (Intel w/o AVX2): synology-opencv-tensorflow)
- Change to new base image: export OpenCV and Tensorflow in new base images (for Raspberry and Synology Version)
- Update roi.jpg on every run
- Workaround for memory leak in tensorflow function predict() - see keras-team/keras#13118
- Reading of analog counters is enabled or disabled via config.ini special thanks to alikanarya
- Error correction in ReadDigitalDigitClass.py
- Upgrade to Tensorflow 2.0
- Changed image processing within CNN to Pillow (instead of OpenCV)
- Training analog and digital CNN with additional type of counter and digits respectivelely
- Correct error for disable Logging
- Update to Pillow 6.2.0
- Impementation of optional consistency check of readout value (not negative, maximum rate)
- Load default configuration, if none is present (beneficial for Docker-Version with mounted config and log directories)
- Parameter "simple" to reduche output to a single value
- Update neural network for readout analog meter
- storage and usage of last full readouts to substitute "NaN" values in digital counters
- The image soucre can be specified within in the config.ini. If no "url"-Parameter is given, the image will be pulled directly from the given source, including logging.
- Update of the name convention of the log-files from Unix-Timestamp to Human-Readable time coding
- Usage of improved analog detection (sinus and cosinus coding)
- New folder structure to support easy implementation in Docker-ContainerUp
- Attention: adaption of INI-File needed, as Python is handling this slightly different
- Implemenation of neural networks with version 1.x (analog and digital) neuron