Issue on page /basic_tutorials/03_detector_vids.html: Long video recording

- For an 18-minute single-person video recording, it takes me an approximated (still running) 30 hours to derive the features w/ the code.
- When I split the video recording into one-minute segment, each segment only takes 10 mins to run. I
- Is it principled for me to use the 2nd approach?
- As I'm trying to derive features for each second of the video, I've made sure skip_frames is set to the # of frames per second (which is 25 in my case).

Start with the workflow shown in /basic_tutorials/03_detector_vids.html and reproduce feature extraction for the 18-minute video and one-minute segments using skip_frames=25. Measure the runtime and determine whether the difference indicates a performance problem or an expected processing constraint; document the principled recommended workflow and any actionable fix.

Written by the indexing model from the issue text.