For PNG input files with multiple levels of transparency (alpha channel values) the output might not show all details or show unexpected effects around shapes. One of the first of steps of the vectorization algorithms is to evaluate each pixel of the input if it is opaque or transparent. Currently no intermediate alpha channel values are supported so intermediate values are rounded to 0% or 100% transparency which might result in these unexpected results. For many images this problem can be easily solved by removing the alpha transparency in the beginning. If for example the transparent background is replaced with white, then all alpha channel values are transformed to opaque gray-level values. Vectorization works best for images without transparent pixels so this most of the times solves the issue.
Identify if the output is missing important details because of an input image with multiple transparency values.
Open the tab 'Transparency' tab and enable Replace transparency with solid color to yes. If a different background color is more adequate for your image you can select it using the color dialog.
Here in this example we also switch to "Sketch" because the input image is only greyscale and has no colors.
All the important details are now visible in the output.
Step 5: (optional)
It is possible to use the "Remove Background" feature to make the background transparent again if required.