From 8b84093ddc7a0195453a697fb49ef85c4cacdb95 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]" This page tries to help you on which model to select.
For more technical/deeper explanations have a look on Neural-Network-Types. For digits on water meters, gas-meters or power meters you can select between two main types of models. This model can recognize full digits. It was the first model version. All intermediate states shown a "N" for not a number. But in post process it uses older values to fill up the "N" values if possible. For digits on water meters, gas-meters or power meters you can select between two main types of models: This model can recognize full digits. It was the first model version. All intermediate states shown a It's possibly a good fallback, if It's possibly a good fallback, if Main features: These models are used to get a continuous reading with intermediate states. To see what the models are doing, you can go to the Recognition page. These models are used to get a continuous reading with intermediate states. To see what the models are doing, you can go to the Recognition page of your device. Main features: Look here for a list of digit images used for the training. The difference is in the internal processing. The dig-class100 is a standard classification model. Each tenth step is an output. dig-cont uses two outputs and arctangent to get the result. You see very complicated. The difference between Note Try both models on your device and take the one that gives you the best results. Look here for a list of digit images used for the training. For pointers on water meters use the analog models. You can only choose between ana-class100 and ana-cont. Both do mainly the same. For pointers on water meters use the analog models: You can choose between two models: Both do mainly the same. Main features: Look here for a list of pointer images used for the training The difference is in the internal processing. Again, the difference between Note Take the one that gives you the best results. Both models learn from the same data. Look here for a list of pointer images used for the training The normally trained network is calculating with internal floating point numbers. The saving of floating point numbers naturally takes more space than an integer type. Often the increased accuracy is not needed. Therefore there is the option, to "quantize" a neural network. In this case the internal values are rescaled to integer values, which is called "quantization". The stored tflite files are usually much smaller and runs faster on the edgeAI-device.
-Usually the models are distrusted therefore in both versions. They can be distinguished by a "-q" at the end of the logfile.
-
-
-
-1.1 Digit Models
-1.1.1 dig-class11
-
+
+dig-class11dig-class100 and dig-cont1.1.1
+dig-class11N for not a number. But in post process it uses older values to fill up the N values if possible.
dig-cont/dig-class100 results are not good.1.1.1.1 Main features
+dig-cont or dig-class100 results are not good.
-ana-class100 or ana-cont)1.1.2 dig-class100 / dig-cont
-1.1.2
+dig-class100 and dig-cont
1.1.2.1 Main features
+
-dig-class11 that results continue to be calculated in the transition between digits.1.1.2.2 dig-class100 vs. dig-cont
-dig-class100 and dig-cont is in the internal processing.
+The dig-class100 is a standard classification model. Each tenth step is an output.
+dig-cont uses two outputs and arctangent to get the result. You see very complicated. 1.2 Analog pointer models
-1.2.1 ana-class100 / ana-cont
-1.2.1
+ana-class100 and ana-cont
1.2.1.1 Main features
+
+
+ana-class100 ana-cont
-1.2.1.2 ana-class100 vs. ana-cont
-ana-class100 and ana-cont is in the internal processing.1.3 Different types of models (normal vs. quantized)
1.3.0.1 Example:
+Usually the models are distrusted therefore in both versions. They can be distinguished by a q at the end of the logfile.
Example: