Teaching Your Laptop to Sniff Out Which Couch You're On

What it does

whereami is a cross-platform Python utility that scans nearby WiFi access points and uses a RandomForest classifier to guess which room—or even which piece of furniture—you’re sitting on. It stores signal fingerprints for labeled locations and compares live scans against that history. The author claims it can distinguish spots roughly two meters apart, like a pair of couches, provided you give it enough varied samples.

The interesting bit

The project is essentially a Python reimplementation of the older find tool, but with the rough edges smoothed out. It ships with its own companion library, access_points, to paper over the mess of cross-platform WiFi scanning, and it turns out vertical height differences are often easier to tell apart than horizontal ones.

Key highlights

• Runs on macOS, Windows, and Linux via a unified WiFi-scanning abstraction.
• Uses sklearn’s RandomForest under the hood; exposes probabilities and cross-validation scores.
• Claims >99% accuracy at ~10 meter distances, and usable granularity down to 2 meters.
• Stores all training data locally in a dotfolder; no cloud dependency.
• Includes a hook for home-automation tinkerers (see the related wherearehue project).

Caveats

• The accuracy section contains a stray keystroke and leans on anecdotal home testing rather than rigorous benchmarks.
• Ultra-short-range predictions risk temporal overfitting unless you alternate sampling locations.
• The README still lists test targets for Python 2.7, 3.4, and 3.5.

Verdict

Worth a look if you want a quick, offline indoor-positioning hack for home automation or local scripting. Skip it if you need production-grade IPS or if your environment lacks a handful of stable WiFi access points.