SC7000 Series Smart Camera User Manual
30
Steps
1. Double click the Deep Learning Tool to run it.
2. Select SC Platform as Target Platform, select Character Training as Training Type, and click
Next.
Note
The SC platform currently only supports 2 training types, including character training and image
classification.
3. Click to add a single image or click to add an image folder.
4. Click to select characters in the image or click to select single character, and enter
characters in the label box.
5. Repeat the previous step to train more samples, and click Next.
6. Set Training Tool Type, Training Parameters, and Save parameters according to actual
demands.
Note
● Character Model Type has 3 types, including single character recognition, text positioning, and
text line recognition. Single character recognition and text line recognition are used to train the
recognition ability, while text positioning is used to train the location ability. You need to train
them separately. Here we take the single character recognition as an example.
● When selecting Single Character Recognition or Text Positioning as Character Model Type, you
can set Max Iteration Times and Basic Learning Rate. Max Iteration Time refers to the times of
character training. You can increase its value to improve train effect, but it will cost more time.
Basic Learning Rate determines the training speed.
● When selecting Text Line Recognition as Character Model Type, you can set Max Iteration
Times, Accuracy Threshold, and Pre-training Model. Accuracy Threshold determines when to
stop the train process, and it is recommended to set it as 0.98 and above. Pre-training Model is
reserved, and you do not need to set it.
To Next Page
Loading...
