Title: Object recognition using shading
1Object recognition using shading
2Object recognition using shading
- Strong claim 1 to recognize an object using
shading, we reconstruct the objects 3D shape
and then we recognize this 3D shape.
3Object recognition using shading
- Strong claim 2 shape from shading is not
used in object recognition
4What is the role of shape from shading in object
recognition ?
- To what extent do humans perceive 3D shape from
shading ? - To what extent do humans use 3D shape perception
to recognize objects ? - We need to answer Q1 before we can answer Q2.
5Human perception of local shape from shading
under variable lighting
- Michael S. Langer Heinrich H. Bülthoff
- Max-Planck-Institute for Biological
CyberneticsTübingen, Germany - McGill University, Montreal, Canada
6Illumination models
- Sunny day (Horn 70, )
- Cloudy day (Langer and Zucker ICCV 93,
Stewart and Langer CVPR 97)
7Overview
- Experiment 1 SFS on a sunny day
- Experiment 2 SFS on a cloudy day
8 SFS on a Sunny Day
L
N(x)
I(x) r N(x) L
9 Depth-reversal ambiguity on a sunny day
hill
valley
10 - Hollow Mask Illusion (Luckiesh, 1916)
11Hollow mask illusion is due to two
factors(Johnston et al 92 , Hill and Bruce
94)
- face vs. hollow mask
- familiarity global
convexity - (recognition)
-
12Our experiments use unfamiliar surfaces
13 Procedure
14 15 Task hill or valley ?
16Three factors were tested
- light source direction (above gt below)
- global shape (convex gt
concave) - viewpoint (above gt
below) -
171. light source direction
- light
- from
- above
- light
- from
- below
-
18 2. global shape
193. Viewpoint (Reichel Todd 1990)
- from above from
below - (floor)
(ceiling)
20Factor 3 viewpoint
-
- view from above
-
- view from below
-
21Factor 3 viewpoint
-
- view from below
-
- view from above
-
22Design
- three factors (2 x 2 x 2)
- - light direction (above, below)
- - global shape (convex, concave)
- - viewpoint (floor, ceiling)
- 512 trials (64 x 8 conditions)
23Results (linear regression)
- percent correct (hill or valley ?)
- 51 10 light direction
- 13 global shape
- 11 viewpoint
- (Each factor had value of 1 or
1)
24Conclusions Experiment 1
- many factors play a role in resolving the
depth-reversal ambiguity - (light direction, global shape, viewpoint,
.)
25Conclusions Experiment 1
- many factors play a role in resolving the
depth-reversal ambiguity - (light direction, global shape, viewpoint,
.) - observers often ignore available image
information (perspective, occluding contours,
shadows)
26Experiment 2 Shape from shading on a cloudy day
27Shading on a cloudy day
?
(x)
28Shading on a Cloudy Day
?
?
I(x)
N(x)
L d L
?
(x)
?
?(x) angle of visible light source
29Shading in a valley on a cloudy day
local intensity maxima
30Local intensity maxima in valleys
31Experiment 2
How well do humans perceive shape from shading in
the presence of these local intensity maxima ?
32Experiment 2
How well do humans perceive shape from shading in
the presence of these local intensity maxima
? Hypothesis (shape from shading
skeptic) Humans use dark means deep model.
33 Procedure
34(No Transcript)
35Task Which is higher ?
36Two conditions
???intensity
correlated
anti-correlated
_
???height
_
anti-correlated
correlated
37Which is higher ?
N17
overall score 65 (above chance)
percent correct
_
38Conclusion Experiment 2
- Dark means deep is too simple a model to
explain human perception of shape from shading
on a cloudy day.
39Big Picture What role does 3D shape perception
play in 3D object recognition?
40Big Picture What role does 3D shape perception
play in 3D object recognition?
- To what extent do we perceive 3D shape ? Were
on the way to answering this question.
41Big Picture What role does 3D shape perception
play in 3D object recognition?
- To what extent do we perceive 3D shape ? Were
on the way to answering this question. - To what extent do we use these 3D shape percepts
to recognize objects ? - The answer to Q2 depends on Q1.
42Computer vision psychophysics
I(x) q(x)
(Langer and Zucker 93)
?
I(x) N(x) L d L
?
?(x)
(Stewart and Langer 96)
43PsychophysicsHuman vs. Computer
1
0.8
q q, N
0.6
percent correct
0.4
0.2
-
0
-
-
human computer
44Conclusions Experiment 2
- Dark means deep is too simple to explain human
perception of shape from shading on a cloudy day - Computing local shape in valleys is an inherently
difficult computational problem on a cloudy
day
45Which point is brighter ?
height
N10
brightness
percent correct
_
Correlation
46 Examples
87 (best)
15 (worst)