The early stages of crystallization of tetrahedral systems remain largely unknown, due to experimental limitations in spatial and temporal resolutions. Computer simulations, when combined with advanced sampling techniques, can provide valuable details about nucleation at the atomistic level. Here we describe a computational approach that combines the forward flux sampling method with molecular dynamics, and we apply it to the study of nucleation in supercooled liquid silicon. We investigated different supercooling temperatures, namely, 0.79, 0.86, and 0.95 of the equilibrium melting point T(m). Our results show the calculated nucleation rates decrease from 5.52+/-1.75x10(28) to 4.77+/-3.26x10(11) m(-3) s(-1) at 0.79 and 0.86 T(m), respectively. A comparison between simulation results and those of classical nucleation theory shows that the free energy of the liquid solid interface gamma(ls) inferred from our computations differ by about 28% from that obtained for bulk liquid solid interfaces. However the computed values of gamma(ls) appear to be rather insensitive to supercooling temperature variations. Our simulations also yield atomistic details of the nucleation process, including the atomic structure of critical nuclei and lifetime distributions of subcritical nuclei.