LANVAR
← The shelf

Clever Cut

Did its thing

A smart 3D trimmer: scan your head, paint zones with different guard lengths, and the blade adjusts itself in real time as you move it.

✳ BEng capstone Embedded CKalman filtergRPCNext.js 2025
✳ How it actually works
3 MM 6 MM 12 MM the head, as scanned — zones painted once in the dashboard LIVE READOUT ZONE FADE — BACK SIDES CROWN FADE → SIDES → CROWN GUARD 3 MM 6 MM 12 MM 3 → 6 → 12 MM BLADE HEIGHT POSE · X/Y 1–3 CM · Z SUB-MM the paint decides — you just move the trimmer
FIG 1 · the trim, animated — paint the zones once, the guard follows
THE DEVICE IMU orientation · I²C UWB radios position · SPI VL53L0X ToF height off the scalp real-time threads simultaneous acquisition, low-latency tracking Kalman fusion custom filter → head pose 1–3 cm X/Y (raw sensors: ~10 cm) · sub-mm Z region map painted zones → guard lengths zone under the blade right now blade motor adjusts mid-cut THE APP gRPC backend streams live pose Next.js dashboard watch your head in 3D · paint the zones you paint here, the trimmer obeys up there ↗
FIG 2 · the machinery — sensors → fusion → blade, with the dashboard riding alongside
The story

One year of R&D with Abhigyan Praveen, Saviru Perera, and Daiyan Islam, and the question was simple: why does a trimmer treat every part of your head the same? Clever Cut scans your head in 3D, lets you “paint” regions on the model and give each one its own guard length, then adjusts the blade zone-by-zone while you cut.

The hard part is knowing where the trimmer is. IMU and ultra-wideband data stream in simultaneously over I²C and SPI on real-time threads, and a custom Kalman filter fuses them down to 1–3 cm of X/Y accuracy — versus roughly 10 cm if you take the sensors at their word. A VL53L0X time-of-flight sensor handles the Z axis at sub-millimetre precision, because a haircut is not the place for rounding errors.

A gRPC backend feeds a Next.js dashboard where you watch your own head in live 3D and set up the regions. Embedded firmware, real-time sensor fusion, and a full web stack converging on one very personal robot — that’s the whole degree in one project.

Next on the shelf Scenery →