LogiCare - Second prize winner at HackZurich'23

By Henri, Ilia, Lune and Jiayao for Logitech challenge

Link to Devpost

Background processes

1. tracking.py: tracks mouse movements, continuously writes them into a csv.
2. blink_yawn_detection.py: taps into web camera feed, recognizes the face, eyes, blinking and yawning.
3. analysis.py: combines results from first two, uses rules and formulas to decide whether to issue a nudge or not.
4. gui/app.py: frontend for data visualization, uses Dash.

Data

• We provide two dumps of data to play with (Ilia's, almost a day of mouse tracking; Henri's, a few hours)
• Raw datapoints are sometimes as frequent as 100 times per second, but there are gaps in data - those are times when the person didn't use the laptop.
• In analysis.ipynb you can find a notebook with analysis of that data, with some graphs. Trajectories are computed from raw data to better grasp a purposeful movement of a mouse; average speed and accuracy are measured (see Banholzer et al., 2021)

Limitations

• There are known issues with mouse tracking on Macs. Meaningful insights will only be produced after around 30 minutes of data.
• Data visualization part still contains mock data, even though real data is accessible. If one so desires, it is possible to modify gui/app.py in order to visualize findings.

How to run?

• pip install requirements.txt
• python app.py
  • this command also triggers run of 4 background processes, which you can also run individually, if you wish.