Split RNA-seq bam by strand: normalize before or after split?

Question

[kalavattam]

When splitting a bam from an RNA-seq experiment by strand for visualization (that is, splitting it into two separate bw files, one for the forward strand, the other for the reverse strand), should the bam be normalized (for example, using one of the normalization options provided by, say, deepTools' bamCoverage --normalizeUsing) before or after the split?