Transcript
Page 1: Scenes From Video Workshop Talk

What’s so good about pieces, Lego and understanding?

Anton van den Hengel

Australian Centre for Visual Technologies (ACVT)The University of AdelaideSouth Australia

Page 2: Scenes From Video Workshop Talk

People think in 3D

Page 3: Scenes From Video Workshop Talk

It has been a theme …

"the perception of solid objects is a process which can be based on the

properties of three-dimensional transformations and the laws of nature”

Larry Roberts (1965)

Page 4: Scenes From Video Workshop Talk

Geometry is not enough

Page 5: Scenes From Video Workshop Talk
Page 6: Scenes From Video Workshop Talk

Structure and semantics interact

Page 7: Scenes From Video Workshop Talk

Structure and geometry interact

Page 8: Scenes From Video Workshop Talk

WHY PLANTS ARE LIKE LEGO

Page 9: Scenes From Video Workshop Talk

Developmental changes in response to drought

Boris Parent, ACPFG

0

1000

2000

3000

4000

5000

6000

7000

30 35 40 45 50 55 60 65

Ab

solu

te g

row

th r

ate

[m

m2

d-1

]

Time after sowing [d]

drought

well watered

39 d after sowing

46 d after sowing

The escape response of Clipper under drought is reflected in

an earlier time of absolute maximum growth

Page 10: Scenes From Video Workshop Talk

Morphological changes in response to drought

Boris Parent, ACPFG

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3

30 40 50 60

Re

lati

ve r

atio

of

sho

ot

are

a /

he

igh

t

Time after sowing [d]

The reduced number of tillers under drought is

reflected in the area/height ratio

Barley cv Clipperdrought

well watered

Page 11: Scenes From Video Workshop Talk

Deep reasoning

• Try to explain as much as possible

• Fine-grained and detailed

• Deep semantics

• And the implied constraints

• Shape is only an intermediate step

Page 12: Scenes From Video Workshop Talk

Deconstruction

Page 13: Scenes From Video Workshop Talk

Silhouettes

• We’re only interested in shape (at least for now)

Page 14: Scenes From Video Workshop Talk

Deconstruction

• Render all possible building blocks in every possible position, and recover its silhouette

• Then reconstruct object silhouettes from templates

• Requires enough camera information to achieve this

Page 15: Scenes From Video Workshop Talk

Template shapes

• nTemplates = nShapes x nPositions x nRotations

• So there are lots of them

• But they are sparsely used

Page 16: Scenes From Video Workshop Talk

Sparse recovery

• \alpha a vector of binary template coefficients

• \Pi a matrix with one template silhouette per column

• y the silhouette of the shape to be recovered

• NP hard and fragile

Page 17: Scenes From Video Workshop Talk

Sparse recovery – L_1 norm

• But there may still be millions of templates, and they’re enormous (|Pixels| x |Images|)

Page 18: Scenes From Video Workshop Talk

Sparse recovery – Random projections

• Random projection by DxS matrix \Phi

• D << S

• \Phi is sparsely sampled from N(0,1)

• But there are still too many templates

Page 19: Scenes From Video Workshop Talk

Sparse recovery - Cropping

• Eliminate templates with a footprint that extends significantly beyond that of the object

• Reduces the number of templates by at least an order of magnitude

• Down to tens to tens of thousands of templates

Page 20: Scenes From Video Workshop Talk

Binarising the solution

• Solutions are not binary

• Randomly generate binary hypotheses from non-binary \alpha

• Evaluate using an accurate composition model

Page 21: Scenes From Video Workshop Talk

Results

Page 22: Scenes From Video Workshop Talk

Results

Page 23: Scenes From Video Workshop Talk

Results

Page 24: Scenes From Video Workshop Talk

Results

Page 25: Scenes From Video Workshop Talk

Results

Page 26: Scenes From Video Workshop Talk

Plants

Page 27: Scenes From Video Workshop Talk

Results

200 400 600 800 1000

0.6

0.7

0.8

0.9

Number of Templates

Fra

ctio

n o

f T

rue

Lea

ves R

eco

vere

d

Max

Search

Viable

Page 28: Scenes From Video Workshop Talk

Results

0 0.01 0.02 0.03 0.04 0.05 0.060

0.02

0.04

0.06

0.08

Noise Level (Fraction of Pixels Changed)

Fra

ctio

n o

f P

ixe

ls E

xp

lain

ed

Max

Search

Page 29: Scenes From Video Workshop Talk

Composition problems

Not a true model of silhouette formation

So doesn’t deal well with template overlap

Working on this by subtracting overlaps, graph-based approaches

Somewhat overcome by…

Page 30: Scenes From Video Workshop Talk

Inequality

• Isn’t physically accurate for foreground pixels, so split

• Background (0) pixels

• And foreground pixels

Page 31: Scenes From Video Workshop Talk

Practicality again

• Only interested in the number of pixels outside the object silhouette, not the location

• So not

• but

Page 32: Scenes From Video Workshop Talk

Practicality again

• Want to ensure that

• Need to project to a lower dimension

• But \Phi_I must have only positive elements

Page 33: Scenes From Video Workshop Talk

A better model of composition

• Left with

Page 34: Scenes From Video Workshop Talk

Constraints - Intersection

Page 35: Scenes From Video Workshop Talk

Constraints - Intersection

• Form J where every row represents a constraint

• If templates i and k intersect then insert a row in J with only elements i and k set to 1

Page 36: Scenes From Video Workshop Talk

Constraints - Support

• Form K where every row represents a constraint

• If template i needs support t set K_ii = t

• If template j provides s support to j then K_ij = -s

Page 37: Scenes From Video Workshop Talk

Measurement benefit tails off

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

0.4

0.5

0.6

0.7

0.8

0.9

1

Noise level (added to camera extrinsics)

Accu

racy (

fra

ctio

n o

f tr

ue

blo

cks r

eco

ve

red

)

Accuracy vs noise for varying numbers of measurements

49

441

1225

2401

3969

5929

8281

11025

Page 38: Scenes From Video Workshop Talk

Results

Page 39: Scenes From Video Workshop Talk

Results

Page 40: Scenes From Video Workshop Talk
Page 41: Scenes From Video Workshop Talk
Page 42: Scenes From Video Workshop Talk

Limitations

• One template per value per parameter

• Fixable?


Top Related