activitywatch / activityscraper

Scrapes the activity the user is experiencing

ActivityScraper

Scrape everything, leave nothing behind.

About

Alternative names: Replaymember, Selfpast

If ActivityScraper is given audio and video recordings, it can annotate them and provide playback of them. We can combine this with ActivityWatch and that enables annotated recordings of all computer activity. Using this we can build a complete story of what happened.

But perhaps more amazingly: we could train ML/AI with the data (since it's labeled) to auto-annotate (using image classifiers and deep learning).

This is an application enabled by ActivityWatch.

Issues

• Storage
  • Lowering resolution and quality (the most basic solution)
  • Can be mediated using a high compression rate (but not an option since it needs to be real-time and energy efficient)
• Work
  • When is the right time? Probably not a good time for quite a while, but cool stuff.

PoC

• Doesn't do recording itself
• Just annotates given screen and/or audio recordings.
• Makes those recordings browsable/viewable in some rudimentary fashion

Future

• Nice playback of the recordings
• Might be able to do recording itself if nice libraries are available
• Try to build auto-annotators using AI/ML (sounds like a possible task for deep learning)

Uses

• Archive and label playing using annotations (would work against such horrors as WebDRM)
  • You would have a searchable copy of every bit of video and audio you played on your computer.

Far Future

• Maybe it could even understand VR content one day.