1. In Workspace Explorer open workspace Examples and in Film strip window select OcrReadNumbers dataset. Drag the Image channel to the ACQUIRE section.

2. Add the ExtractText filter. Connect the outImage from the previous filter to the inImage of the current one.

3. Click on the ExtractText filter and in the Properties window select the inRoi to specify the area where characters should be extracted from:

4. Now choose the inSegmentationModel input and create a proper segmentation model to correctly extract characters:

5. If you successfully extracted characters, add the ReadText. Connect the outCharacters output from the previous filter to the inCharacters input of the current one.

6. Click on the ReadText filter and choose the inOcrModel. In the OCR Model Trainer you need to specify the value of each extracted character. When you are done, click Add Samples and Train:

7. Go to the Details tab where for each character you can create variants by opening Variation Toolbox. Please note that you need to select a character first:

8. You can adjust parameters of your training as well as shown in the picture below. If you are ready, you can train your model by clicking on the Train button. When it is over, the training accuracy will be computed:

9. Now you can leave the editor and go back to your program. Your OCR application is ready and if some improvements are necessary, it is possible to add new samples or make suitable changes.

マクロフィルタ メイン