Low-FidelityVideoEncoderOptimizationfor TemporalActionLocalization

This is why the two-stage optimization pipeline as described above becomes the most common andfeasible choice inpracticeforoptimizing aTALmodel.