• It's dependent on how the libraries were made, so just ask whomever made them (having said that, read#2 is commonly determines the strand).
• Just look at a couple genes in IGV. If you color reads by the orientation of read#1 then it should be quickly apparent which (if either) read is determining the strand.

Quick BLAST must do this. Take subset of reads. Blast it again few reference genes and see how your forward reads and reverse reads are aligning by looking at the subject start and subject end. Check if most of your reads are aligning (-) orientation or in (+) orientation.

You could also use infer_experiment.py from the RSeQC package. It will tell you to which strand your read map (and if it is the second or first read) or if your library is not strand specific.