Magic Mirror

02 | Devlog

Here is my MagicMirror.

Personal profiles are live, the Obsidian vault is linked, and the mirror knows who’s in front of it.

what I did:

• built out personal profiles with custom profile photos and rounded borders
• profile switching is fully automatic and activated via facial recognition
• after 20 seconds of no face detected, the mirror drops back to lock mode
• if the PIR also hasn’t triggered in those 20 seconds, it goes full sleep
• linked the mirror to my local Obsidian vault by parsing the daily markdown file
• daily objectives from my notes now show up live on the display

how the profile switch works:

• the moment it picks up your face, your personal modules load in
• once you walk away, the 20 second timer starts.
• if there is no face and no motion then the mirror sleeps
• it’s a two-layer falloff that keeps the mirror from staying lit for no reason so that I can sleep at night

Obsidian integration:

I’m pulling from my daily notes vault by reading and parsing the markdown file directly on the pi. whatever objectives I’ve written for the day show up on the mirror automatically through my private github repository

current stats:

• profiles: 2 (Brady + Camren)
• profile trigger: face recognition, automatic
• lock timeout: 20 seconds after last face detected
• sleep trigger: no PIR + no face for 20 seconds
• Obsidian source: local daily notes vault (syncing with github), parsed from markdown

what’s next:

• implementing the SR-04 ultrasonic sensor
• soldering the modules
• CADing housing for the mirror modules

01 | Devlog

Here is my MagicMirror.

I successfully cloned the magicmirror repo to my raspberry pi, and now my phone’s apple calendar is finally showing up on the display. I plan to later implement facial recognition and integration into my obsidian vault.

what I did:

• cloned the magicmirror repo and installed it on my pi
• configured the calendar to pull from my personal apple calendar
• time and weather displayed
• got facial recognition working on my logitech usb camera

current stats:

• Pi model: 4B with 4GB ram
• Webcam: Logitech Pro Stream Webcam
• mirror UI: 120 fps
• facial recognition frames: 15 fps
• CV scalar: 1

I decided to go with a CV scalar of 1 to improve the model’s scanning at longer distances since the pi held up under the heavier processing.

what’s next:

• I plan to add personal profiles based on the facial recognition model
• each profile will have their personalized calendar and sports teams’ schedule displayed
• then I’ll be adding a SR-501 InfraRed sensor and a SR-04 Ultrasonic sensor to wake up the mirror from sleep mode