Unfolding an Indoor Origami World David Fouhey, Abhinav Gupta, - - PowerPoint PPT Presentation

unfolding an indoor
SMART_READER_LITE
LIVE PREVIEW

Unfolding an Indoor Origami World David Fouhey, Abhinav Gupta, - - PowerPoint PPT Presentation

Unfolding an Indoor Origami World David Fouhey, Abhinav Gupta, Martial Hebert 1 2 3 Local Evidence Hoiem et al. 2005, Saxena et al. 2005, Fouhey et al. 2013, etc. 4 Constraints 5 Constraints for Single Image 3D Local Smoothness Low


slide-1
SLIDE 1

Unfolding an Indoor Origami World

David Fouhey, Abhinav Gupta, Martial Hebert

1

slide-2
SLIDE 2

2

slide-3
SLIDE 3

3

slide-4
SLIDE 4

Local Evidence

4

Hoiem et al. 2005, Saxena et al. 2005, Fouhey et al. 2013, etc.

slide-5
SLIDE 5

Constraints

5

slide-6
SLIDE 6

Constraints for Single Image 3D

6

Local Smoothness

Low Level, Generic

Hoiem et al. 2005, Saxena et al. 2005, 2008, Munoz et al., 2009, etc.

slide-7
SLIDE 7

Constraints for Single Image 3D

7

Local Smoothness

Low Level, Generic

Hoiem et al. 2005, Saxena et al. 2005, 2008, Munoz et al., 2009, etc.

slide-8
SLIDE 8

Constraints for Single Image 3D

8

Low Level, Generic High Level, Physical

slide-9
SLIDE 9

High Level, Physical

Constraints for Single Image 3D

9

Local Smoothness

Low Level, Generic

Coughlan and Yuille 2000, etc.

slide-10
SLIDE 10

High Level, Physical

Constraints for Single Image 3D

10

Local Smoothness

Low Level, Generic

Hedau et al. 2009, Del Pero et al., 2011, Wang et al., 2012, Schwing et al. 2012, etc.

slide-11
SLIDE 11

High Level, Physical

Constraints for Single Image 3D

11

Local Smoothness

Low Level, Generic

Lee et al. 2010, Xiao et al. 2012, Zhao et al. 2013, Schwing et al., 2013, etc.

slide-12
SLIDE 12

High Level, Physical

Constraints for Single Image 3D

12

Low Level, Generic

slide-13
SLIDE 13

Mid-level in the Past

13

Huffman 71, Clowes 71, Kanade 80, 81 Sugihara 86, Malik 87, etc.

slide-14
SLIDE 14

Our Mid-Level Constraints

14

slide-15
SLIDE 15

This Work

15

Input: Single Image Output: Discrete Scene Parse

slide-16
SLIDE 16

Overview

16

Parameterization Formulation Experimental Results

slide-17
SLIDE 17

Overview

17

Parameterization Formulation Experimental Results

slide-18
SLIDE 18

Parameterization

18

slide-19
SLIDE 19

Parameterization

19

vp1 vp2 vp3

VP Estimator from Hedau et al., 2009

slide-20
SLIDE 20

Parameterization

20

Two VPs give grid cell

slide-21
SLIDE 21

Encoding Surface Normals

21

slide-22
SLIDE 22

Encoding Surface Normals

22

slide-23
SLIDE 23

Encoding Surface Normals

23

slide-24
SLIDE 24

Encoding Surface Normals

24

x1,…, x400 x401,…, x800 x801,…, x1200

slide-25
SLIDE 25

Related Parameterizations

25

vp1 vp3

x2 x1 x3 x4

Hedau et al., 2009; Wang et al. 2010, Schwing et al., 2012, 2013

vp2

slide-26
SLIDE 26

Overview

26

Parameterization Formulation Experimental Results

slide-27
SLIDE 27

Parameterization

27

slide-28
SLIDE 28

Formulation

28

slide-29
SLIDE 29

Unaries

29

slide-30
SLIDE 30

Unaries

30

Low c Any 3D Evidence High c

slide-31
SLIDE 31

Unaries

31

Input 3D Primitive Bank

Local: Data-Driven 3D Primitives

Fouhey, Gupta, Hebert, 2013

slide-32
SLIDE 32

Unaries

32

Hedau, Hoiem, Forsyth, 2009

Global: Cuboid Fit + Clutter Mask

Input Predicted Walls Clutter Mask

slide-33
SLIDE 33

Binaries

33

slide-34
SLIDE 34

Convex/Concave Constraints

34

Convex (+) Concave (-)

slide-35
SLIDE 35

Convex/Concave Constraints

35

Detected Concave (-)

slide-36
SLIDE 36

Convex/Concave Constraints

36

Detected Concave (-)

slide-37
SLIDE 37

Convex/Concave Constraints

37

Detected Concave (-)

slide-38
SLIDE 38

Convex/Concave Constraints

38

Detected Concave (-)

slide-39
SLIDE 39

Convex/Concave Constraints

39

Detected Concave (-)

slide-40
SLIDE 40

Detecting Convex/Concave

40

Ground-Truth Discontinuities similar to Gupta, Arbelaez, Malik, 2013 3DP from Fouhey, Gupta, Hebert, 2013

Input 3D Primitive Bank

Use 3DP to Transfer Discontinuities

slide-41
SLIDE 41

Smoothness

41

slide-42
SLIDE 42

Constraints

42

slide-43
SLIDE 43

Solving the Model

43

slide-44
SLIDE 44

Overview

44

Parameterization Formulation Experimental Results

slide-45
SLIDE 45

Dataset

NYU Depth v2: 795 Train, 654 Test

45

slide-46
SLIDE 46

Qualitative Results

46

slide-47
SLIDE 47

Qualitative Results

47

slide-48
SLIDE 48

Qualitative Results

48

slide-49
SLIDE 49

Qualitative Results

49

slide-50
SLIDE 50

Surface Connection Graphs

50

Convex Concave

slide-51
SLIDE 51

Baseline

Primary Baseline: 3D Primitives

51

Fouhey, Gupta, Hebert, 2013

Output Input

slide-52
SLIDE 52

Quantitative Results

52

Summary Stats (⁰) (Lower Better) % Good Pixels (Higher Better)

11.25⁰ 22.5⁰ 30⁰ Mean Median

3DP 35.9 52.0 57.8 36.0 20.5 49.4 Hedau et al. 34.2 49.3 54.4 40.0 23.5 54.1

RMSE

Lee et al. 18.6 38.6 49.9 43.3 36.3 54.6 Proposed 37.6 53.3 58.9 35.1 19.2 48.7

slide-53
SLIDE 53

Quantitative Results

53

slide-54
SLIDE 54

Failure Modes

Mistaken but Confident Evidence

54

slide-55
SLIDE 55

Failure Modes

55

Missing High-Level Modeling

slide-56
SLIDE 56

Conclusions

56

Single Image Parameterization Formulation Discrete Parse

slide-57
SLIDE 57

Thank You

57