- Additional images, new type of digits
- Updated quantization implementation (use original images)
- Internal improvements (speed)
- Update with additional set of digits
- Update with additional set of digits
- Removing of double images
- Adding of V200 digits
- NEW 9.0.1: Additional digits for the last variant
- New type of LCD-numbers integrated (white on black)
- Update with additional set of digits
- Update with additional set of digits
- Update with additional set of digits
- Update with additional set up digits
- Update with additional set up digits
-
Update with additional set up digits
-
Change file naming convention (tflite file)
- Update with additional set up digits
- Update with additional set up digits
- Update with additional set up digits
- Included LCD-digits in training
- Update with additional set up digits (gas counter, electric power counter)
- Update with additional set up digits
- Relabeling of training images
- Take out unambiguous (number or NaN) - criteria for number: complete picture is in the ROI visible Labeling-Criteria.md
- Remove replicated pictures
- Implementation of "train all" parameter
- Experiments with reduces network size for speeding up image recognition ("...-Small-vX.tflite")
- Training with new picture from iobroker users
- White on black, white on red digits - additional pictures
- Training with new picture from iobroker users
- White on black, white on red digits
- Training with new picture from iobroker users
- Training with new picture from iobroker users
- Training with new picture from iobroker users
- For size minimized tflite-File implemented ("_quantized")
- Updated to Tensorflow 2.1
- additional export to TF-Lite Version (.tflite)
- Training with new picture from iobroker users
- Training with new picture from iobroker users
- Removal of standalone server - (included in main project)Training of additional digital number (provided from iobroker users)
- Training of additional digital number (provided from iobroker users)
- Image processing changed to Pillow (remove OpenCV)
- Usage of Tensorflow 2.0 for training
- Change Image handling completely to OpenCV-Library
- Learning with increased ZoomRange (10% --> 40%), learning within one step