Using SlowFast with short video sequences

I'm working on a project to detect rear turning signals and brake lights in vehicles. Most of the sequences have a length of around 16 frames. I saw in the model zoo, the lowest pre-trained frame length x sample rate is 4 x 16. From the paper, I believe frame length is the number of frames used by the Slow sequence, and the sample rate is the temporal stride. Therefore, this makes me think that the lowest number of frames this pre-trained model can work with is 4 * 16 = 64 frames.

Is it possible to use the pre-trained 4 x 16 model with sequences that are so short (~16 frames)?

Thanks!