Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 48

What's This

Glove?
Oh…
Passive Stylus
• Thick pen-tip
Active Stylus
• Need recharging
Ideal Stylus?
Passive
+
Accurate
DodecaPen
Dodecahedron + Pen

6DoF Tracking
How To Use?
Related Work
Lumitrack
(UIST 2013) Accuracy: 5mm
IrPen
(CGA 2014)

Accuracy: 10mm
Light Chisel
(PG 2015)

Accuracy: 2mm
DodecaPen
(Proposed)

Accuracy: 0.4mm
How To Implement?
Proposed 6DoF Pose Tracking System

Inter-frame Corner
Tracking Marker Intensity
+ Normalization

No
Camera DodecaPen
Did APE
Succeed? Digital 2D Drawing
Yes

Approximate Pose Estimation Dense Pose Refinement


Input Frames (APE) Output Pen-tip Trajectory
(Based on DodecaPen Poses)
Why Dodecahedron?
Pose Jumping!
Multiple Candidates due to Coplanar Points
Platonic Solid
A Regular, Convex Polyhedron

Tetrahedron Cube Octahedron Dodecahedron Icosahedron


Four faces Six faces Eight faces Twelve faces Twenty faces

(source: wikipedia)
The Chosen One
Ideal ⟺ Real
Fail
Dodecahedron
Calibration
Pen-tip
Calibration
6 DoF Pose Tracking
+ =

+ =?
Performance Analysis
Evaluation with Synthetic Data

24 Motion Patterns
4 Varying Conditions
Shot Noise Spatial Blur Camera Resolution Mask Kernel Width
Rotation
Error
Translation
Error
Success
Rate
Evaluation with Real Data

4 Real Drawings
VS. Mocap System (16 Cameras)
Boba Thumb DodecaPen UIST2017

APE

Proposed

Mocap
DodecaPen VS. Mocap

+ VS. $34,990

Comparable to a mocap system


with 10 active cameras
Take Home Message
1. Sub-millimeter accurate 6DoF tracking
using a set of readily available and easy-
to-assemble components.
2. Single camera pose estimation can be fast
and robust enough for drawing in 2D, 3D
and in VR.
DodecaPen: Accurate 6DoF Tracking of a Passive Stylus
Inter-frame Corner Tracking (ICT)
• Pyramidal Lucas-Kanade marker
corner tracking
• P𝑛P algorithm to get the initial Marker Intensity
pose 𝐩′ Normalization
+

No

Camera DodecaPen Did APE


Succeed?
Yes Digital 2D Drawing
Mipmap Masks
Marker Mipmaps

Masked Mipmaps

Approximate Pose Estimation (APE) Dense Pose Refinement (DPR)


• Marker Detection • Minimize appearance distance with Gauss
• Minimize reprojection error with P𝑛P Newton and backtracking line search
(BLS) to get the final pose 𝐩∗ Output Pen-tip Trajectory
Input Frames algorithm to get the initial pose 𝐩′
• Marker & mask mipmaps
(Based on DodecaPen Poses)

You might also like