4-postprocessing.md

Tutorial 4: Postprocessing

This tutorial shows how to enable the BaseHandwrittenPostprocessor for the BaseHandwrittenModel in order to apply a set of image augmentation filters to the resulting image.

The model by default has postprocessing disabled, since the two used libraries by the postprocessor (Augraphy and Albumentations) require citations to be given when used.

Standard model use

To recap, this is the standard way in which you would use the BaseHandwrittenModel to generate synthetic images:

import cv2
import smashcima as sc

model = sc.orchestration.BaseHandwrittenModel()
scene = model("my-input-file.musicxml")

for i, page in enumerate(scene.pages):
    bitmap = scene.render(page)
    cv2.imwrite(f"page_{i}.png", bitmap)

Replacing the postprocessor

All you need to do is to replace the Postprocessor implementation in the model's services. This is done by inheriting from the base model class and modifying the service registration code:

class MyModel(sc.orchestration.BaseHandwrittenModel):
    def register_services(self):
        super().register_services()
        
        # for postprocessor use the base handwritten one,
        # instead of the default NULL postprocessor
        self.container.type(sc.BaseHandwrittenPostprocessor)
        self.container.interface(
            sc.Postprocessor, # when people ask for this
            sc.BaseHandwrittenPostprocessor # construct this
        )

Then after you run the model, the produced bitmap should have postprocessing filters applied:

Controlling filter stacks

The BaseHandwrittenPostprocessor defines a number of filter stacks. These can be manually disabled if needed:

# get the postprocessor instance
p: sc.BaseHandwrittenPostprocessor = model.postprocessor

# never apply camera effects
p.f_camera.force_dont = True

# and always apply scribbles
p.f_scribbles.force_do = True

Citing Augraphy and Albumentations

If you use the BaseHandwrittenPostprocessor or you build your own using these libraries, don't forget to cite their respective papers:

• Augraphy citation
• Albumentations citation

Conclusion

You've learned how to enable postprocessing on the base model. You can learn more about postprocessing and compositing in the technical documentation of Smashcima.