Via BEDOPS, extract gene features from your GFF file and convert to BED:

$ awk '$3=="gene"' annotations.gff | gff2bed - > genes.bed

Then filter on strand and pass to wc -l to count the number of lines:

$ awk '$6=="+"' genes.bed | wc -l
*number of forward-strand genes*

And:

$ awk '$6=="-"' genes.bed | wc -l
*number of reverse-strand genes*

Converting to BED is not strictly required, but it makes sense if you're doing counting operations, which are essentially set operations. BED is a better format for set operations than GFF.

Here is a bash solution as well:

#First lets define a couple of variables to act as counters for each strand ( + & - )

forward=0
reverse=0

#Now we create a loop with your lines of interest from the gff3 file as the changing variable

for line in $(cat file.gff3 | grep '.*gene.*[[:space:]]+[[:space:]]')
do
  forward=$(expr "$forward" + 1) #this will add 1 to the counter for each line found by grep
done

#Depending on your grep version you might need to scape the "+" character, BSD grep doesn't have to

#Same process but for the reverse strand

 for line in $(cat file.gff3 | grep '.*gene.*[[:space:]]-[[:space:]]')
 do
   reverse=$(expr "$reverse" + 1)
 done

#Finally we print the results

echo -e "\nGenes present in the forward (positive) strand: $forward"
echo -e "Genes present in the reverse (negative) strand: $reverse\n"

Hope this helps.