I have a plolygon shapefile for the changes in the amazon rainforest and want to make use of it to extract segmented masks to use it as a labeled data in my deep learning model the polygon shapefile

I have tried the QGIS to convert it to raster format but I don’t know which pixel size I should use? is it will be the same as the satellite images that Iam gonna use in my training or something different because Iam using (Landsat 8) with only spatial resolution 30m and It will not be perfect for training masks to be the same as the Landsat 8. The training masks with a pixel size 20000
With resolution only 30m as the Landsat 8