131
Non-line-of-sight imaging 15-463, 15-663, 15-862 Computational Photography Fall 2017, Lecture 25 http://graphics.cs.cmu.edu/courses/15-463

See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Non-line-of-sight imaging

15-463, 15-663, 15-862Computational Photography

Fall 2017, Lecture 25http://graphics.cs.cmu.edu/courses/15-463

Page 2: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Course announcements

• Homework 6 will be posted tonight.- Will be due Sunday 10th.- Almost no coding, only data capture and using existing code.- You will need high-sensitivity cameras, so use the DSLRs you have picked up.

• Final project report deadline moved to December 15th.- Originally was December 11th.

Page 3: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Overview of today’s lecture

• The non-line-of-sight (NLOS) imaging problem.

• Active NLOS imaging using time-of-flight imaging.

• Active NLOS imaging using WiFi.

• Passive NLOS imaging using accidental pinholes.

• Passive NLOS imaging using accidental reflectors.

• Passive NLOS imaging using corners.

Page 4: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Slide credits

Many of these slides were directly adapted from:

• Shree Nayar (Columbia).• Fadel Adib (MIT).• Katie Bouman (MIT).

Page 5: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

The non-line-of-sight (NLOS) imaging problem

Page 6: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Time-of-flight (ToF) imaging

Page 7: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Looking around the corner

Page 8: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Looking around the corner

Page 9: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Active NLOS imaging using time-of-flight imaging

Page 10: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Processing a single-pixel transient

light

source

transient camera

single-pixel

time

intensity

VSVD

Page 11: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Processing a single-pixel transient

light

source

transient camera

single-pixel

time

intensity

VSVD

t1 = 10 ps

Page 12: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Processing a single-pixel transient

light

source

transient camera

single-pixel

time

intensity

t1 = 10 ps

lout

lin

lnlos = t1 ∙ c - lin - lout

Where in the world can we find

points creating this pathlength?

VSVD

Page 13: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Processing a single-pixel transient

light

source

transient camera

single-pixel

time

intensity

t1 = 10 ps

lout

lin

lnlos = t1 ∙ c - lin - lout

Ellipse with foci on the virtual source

and detector and pathlength lnlos

VSVD

Page 14: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Processing a single-pixel transient

light

source

transient camera

single-pixel

time

intensity

t2 = 20 ps

lout

lin

lnlos = t2 ∙ c - lin - lout

Ellipse with foci on the virtual source

and detector and pathlength lnlos

VSVD

Page 15: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Processing another single-pixel transient

light

source

transient camera

single-pixel

time

intensity

t2 = 20 ps

lout

lin

VSVD

• What assumption are we making here?

• Is there any other information we are not using?

Page 16: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Processing another single-pixel transient

light

source

transient camera

single-pixel

time

intensity

t2 = 20 ps

lout

lin

VSVD

• What assumption are we making here?

• Is there any other information we are not using?

Photons only interact

with the NLOS scene

once (3-bounce paths)

Page 17: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Processing another single-pixel transient

light

source

transient camera

single-pixel

time

intensity

t2 = 20 ps

lout

lin

VSVD

• What assumption are we making here?

• Is there any other information we are not using?

Photons only interact

with the NLOS scene

once (3-bounce paths)

We need to also

account for intensity

Page 18: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Elliptic backprojection

light

source

transient camera

single-pixel

lout

lin

VSVD

Page 19: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Elliptic backprojection

Page 20: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Elliptic backprojection

Page 21: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Hidden Rooms

Visible wall

Invisible walls

Page 22: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Hidden Rooms: One plane at a time

Visible wall

Invisible walls

Page 23: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Hidden Rooms: One plane at a time

Visible wall

Invisible walls

Page 24: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Hidden Rooms: One plane at a time

Visible wall

Invisible walls

Page 25: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Hidden Rooms: One plane at a time

Visible wall

Invisible walls

Page 26: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Hidden Rooms: One plane at a time

Visible wall

Invisible walls

Page 27: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Hidden Rooms: One plane at a time

Visible wall

Invisible walls

Page 28: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Rectangular Rooms

Original room Reconstructed room

• Average AICP error for all the walls is TBD mm (TBD %). – Normalized with average room

length of 1.1m

Page 29: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Reconstructing Complex Shape and Reflectance

wall

camerahidden object

what depth our camera sees

what a regular camera sees

wall

camerahidden object

what shape our camera sees

what a regular camera sees

Page 30: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Active NLOS imaging using WiFi

Page 31: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Imaging through occlusions

Page 32: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Imaging through occlusionsusing radio frequencies

Page 33: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Key Idea

Page 34: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Challenges

Wall refection is 10,000x stronger than reflections coming from behind the wall

Tracking people from their reflections

Page 35: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Wi-Vi: Small, Low-Power, Wi-Fi

• Eliminate the wall’s reflection

• Track people from reflections

• Gesture-based interface

• Implemented on software radios

Page 36: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

How Can We Eliminate the Wall’s Reflection?

Page 37: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Idea: transmit two waves that cancel each other when they reflect off static objects but not moving objects

Wall is static

People tend to move

disappears

detectable

Page 38: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Eliminating the Wall’s Reflection

Transmit Antennas

Receive Antenna:

αx

x

Page 39: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Eliminating the Wall’s Reflection

h2

h1

y = h1 x + h2αx0

α = -h1 / h2

Receive Antenna:

αx

x

Page 40: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Eliminating All Static Reflections

Page 41: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Eliminating All Static Reflections

y = h1 x + h2αx

Static objects (wall, furniture, etc.) have constant channels

y = h1 x + h2(- h1/ h2)x0

People move, therefore their channels change

y = h1’ x + h2

’(- h1/ h2)x

Not Zero

h2

h1

αx

x

Page 42: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

How Can We Track Using Reflections?

Page 43: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Direction of reflection

θ

Antenna Array

Tracking Motion

RF source

Page 44: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

θ

Antenna Array

Tracking MotionDirection of motion

At any point in time, we have a single measurement

Page 45: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

θ

Antenna Array

Tracking MotionDirection of motion

θ

Direction

of motion

Page 46: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

θ

Antenna Array

Tracking MotionDirection of motion

θ

Direction

of motionHuman motion emulates antenna array

Page 47: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 48: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

A Through-Wall Gesture Interface

• Sending Commands with Gestures

• Two simple gestures to represent bit ‘0’ and bit ‘1’

• Can combine sequence of gestures to convey longer message

Page 49: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Gesture Encoding

Bit ‘0’: step forward followed by step backward

Bit ‘1’: step backward followed by step forward

Step Backward

Step Forward

Page 50: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Bit ‘1’Bit ‘0’

Gesture Decoding

Matched Filter

Peak Detector

Gesture interface that works through walls and none-line-of-sight

Page 51: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Imaging through occlusionsusing radio frequencies

Head

Right Arm

Chest

Left Arm

LowerLeft Arm

LowerRight Arm

Feet

Our output

Camera image

Our RF sensor

Page 52: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

RF ImagingTraditional Imaging

• Cannot image through occlusions like walls

• Form 2D images using lenses

• Walls are transparent and can image through them

• No lenses at these frequencies

Page 53: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Imaging with RF

Antenna

No lens at these frequencies

Antenna cannot distinguish bounces from different directions

Page 54: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

AntennaArray

Extend to 3D with time-of-flight measurements by repeating this at every depth

Beamforming: Use multiple antennas to scan reflections within a specific beam

Imaging with RF

Page 55: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Coarse-to-fine Scan

• Larger aperture (more antennas) means finer resolution

Used antennas

Page 56: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Used antennas

Coarse-to-fine ScanReflectors

Page 57: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Scannedregion

Coarse-to-fine Scan

Used antennas

Page 58: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Scannedregions

Coarse-to-fine Scan

Used antennas

Page 59: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

• Walls are transparent and can image through them

• No lenses at these frequencies

• No reflections from most points: all reflections are specular

RF ImagingTraditional Imaging

• Cannot image through occlusions like walls

• Form 2D images using lenses

• Get a reflection from all points: can image all the body

Page 60: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Output of 3D RF Scan

Blobs of reflection power

Challenge: Don’t get reflections from most points in RF

Page 61: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

At frequencies that traverse walls, human body parts are specular (pure mirror)

Challenge: Don’t get reflections from most points in RF

Page 62: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

An

ten

nas

Cannot Capture Reflection

At every point in time, get reflections from only a subset of body parts

At frequencies that traverse walls, human body parts are specular (pure mirror)

Challenge: Don’t get reflections from most points in RF

Page 63: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Solution Idea: Exploit Human Motion and Aggregate over Time

RF Imaging Array

Page 64: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

RF Imaging Array

Solution Idea: Exploit Human Motion and Aggregate over Time

Page 65: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

3m 2.5m 2m

Chest (Largest Convex Reflector)

Use it as a pivot: for motion compensation and segmentation

Human Walks toward Sensor

Page 66: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

3m 2.5m 2m

Human Walks toward Sensor

Chest (Largest Convex Reflector)

Use it as a pivot: for motion compensation and segmentationCombine the various snapshots

Right arm

Left arm

Head

Lower TorsoFeet

Page 67: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Human Walks toward Sensor

Page 68: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Implementation

• Hardware– 2D Antenna Array

– Built RF circuit• 1/1,000 power of WiFi

• USB connection to PC

• Software– Coarse-to-fine algorithm implemented in GPU to

generate reflection snapshots in real-time

Rx

Tx

Page 69: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Evaluation

• RF-Capture sensor placed behind the wall

• 15 participants

• Use Kinect as baseline when needed

SnapshotsareCollectedacrossTimeSensorisPlacedBehindaWall CapturedHumanFigure

SENSORSETUP CAPTUREALGORITHM

Sensor

Head

RightArm

Chest

Le Arm

LowerLe Arm

LowerRightArm

Feet

Page 70: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Sample Captured Figures through Walls

Page 71: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Sample Captured Figures through Walls

-0.2 0 0.2 0.4 0.6 0.8x-axis (meters)

0

0.5

1

1.5

2

y-a

xis

(m

ete

rs)

-0.2 0 0.2 0.4 0.6 0.8x-axis (meters)

0

0.5

1

1.5

2y-a

xis

(m

ete

rs)

-0.2 0 0.2 0.4 0.6 0.8x-axis (meters)

0

0.5

1

1.5

2

y-a

xis

(m

ete

rs)

-0.2 0 0.2 0.4 0.6 0.8x-axis (meters)

0

0.5

1

1.5

2

y-a

xis

(m

ete

rs)

Page 72: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Tracking result

Page 73: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Writing in the air

Median Accuracy is 2cm

Kinect (in red)

Page 74: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Passive NLOS imaging using accidental pinholes

Page 75: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

What does this image say about the world outside?

Page 76: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Accidental pinhole camera

Page 77: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Accidental pinhole camera

window is an aperture

projected pattern on the wall window with smaller gap

upside down view outside window

Page 78: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Accidental pinspeck camera

Page 79: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Passive NLOS imaging using accidental reflectors

Page 80: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Ophthalmology

• Anatomy, physiology

[Helmholtz 09;…]

Computer Vision

•Iris recognition

[Daugman 93;…]

HCI

• Gaze as an interface

[Bolt 82;…]

Computer Graphics

• Vision-realistic Rendering

[Barsky et al. 02;…]

Page 81: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Corneal Imaging System

Page 82: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Geometric Model of the Cornea

mm5.5mm18.2 Lb

rt

R=7.6mm

eccentricity = 0.5

Cornea

ScleraIris Pupil

Page 83: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Self-calibration: 3D Coordinates, 3D Orientation

Page 84: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Viewpoint Loci

camera pupil

Z

X

cornea

Page 85: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Viewpoint Loci

Z

X

cornea

camera pupil

Page 86: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Viewpoint Loci

cornea

camera pupil

Page 87: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Resolution and Field of View

X

Z

gaze direction

cornea

Page 88: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Resolution and Field of View

X

Z

FOV

gaze direction

cornea

resolution

Page 89: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Resolution and Field of View

X

Z

FOV

gaze direction

cornea

27 45person’s FOV

resolution

Page 90: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 91: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Environment Map from an Eye

Page 92: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

What Exactly You are Looking AtEye Image:

Computed Retinal Image:

Page 94: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 95: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Corneal Stereo System

right cornea left cornea

camera pupil

epipolar curve

epipolar plane

Page 96: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

From Two Eyes in an Image …

Epipolar CurvesReconstructed Structure (frontal and side view)

Y

X

X

Y

Z

Page 97: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Eyes Reveal …

• Where the person is

• What the person is looking at

• The structure of objects

Page 98: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Implications

Human Affect Studies: Social Networks

Security: Human Localization

Advanced Interfaces: Robots, Computers

Computer Graphics: Relighting [SIGGRAPH 04]

Page 99: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Dynamic Illumination in a Video

Page 100: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Point Source Direction from the Eye

intensity

Page 101: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Point Source Trajectory from the Eye

(deg.) (deg.)

(deg.)

(deg.)

Page 102: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Inserting a Virtual Object

Page 103: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Sampling Appearance using Eyes

Page 104: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Sampling Appearance using Eyes

Page 105: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Computed Point Source Trajectory

intensity

Page 106: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Fitting a Reflectance Model

Page 107: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

3D Model Reconstruction

Page 108: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Relighting under Novel Illumination

Page 109: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

VisualEyesTM

http://www.cs.columbia.edu/CAVE/

with Akira Yanagawa

Page 110: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

Passive NLOS imaging using corners

Page 111: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 112: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 113: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 114: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 115: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 116: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 117: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 118: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 119: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 120: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 121: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 122: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 123: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 124: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 125: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 126: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 127: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 128: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 129: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 130: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:
Page 131: See Through Walls with Wi-Fi - Computer Graphicsgraphics.cs.cmu.edu/courses/15-463/2017_fall/lectures/lecture26.pdf• No lenses at these frequencies • No reflections from most points:

References

Basic reading:• Kirmani et al., “Looking around the corner using ultrafast transient imaging,” ICCV 2009 and

IJCV 2011.• Velten et al., “Recovering three-dimensional shape around a corner using ultrafast time-of-

flight imaging,” Nature Communications 2012.the two papers showing how ToF imaging can be used for looking around the corner.

• Abib and Katabi, “See Through Walls with Wi-Fi!,” SIGCOMM 2013.• Abib et al., “Capturing the Human Figure Through a Wall,” SIGGRAPH Asia 2015.

the two papers showing that WiFi can be used to see through walls.• Torralba and Freeman, “Accidental Pinhole and Pinspeck Cameras,” CVPR 2012.

the paper discussing passive NLOS imaging using accidental pinholes.• Nishimo and Nayar, “Corneal Imaging System: Environment from Eyes,” IJCV 2006.

the paper discussing passive NLOS imaging using accidental reflectors.• Bouman et al., “Turning corners into cameras: Principles and Methods,” ICCV 2017.

the paper discussing passive NLOS imaging using corners.

Additional reading:• Pediredla et al., “Reconstructing rooms using photon echoes: A plane based model and

reconstruction algorithm for looking around the corner,” ICCP 2017.the paper on NLOS room reconstruction using ToF imaging.

• Nishimo and Nayar, “Eyes for relighting,” SIGGRAPH 2004.a follow-up paper to the paper on corneal imaging, show how similar ideas can be used for relighting and other image-based rendering tasks.