MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
Transcript of MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 1/97
Lecture 3Levels of categorization
and multiclass object recognition
6.870 Object Recognition and Scene Understandinghttp://people.csail.mit.edu/torralba/courses/6.870/6.870.recognition.htm
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 2/97
Class business
Start thinking about the class project
Choose a partner (if you want to do it alone, thatis fine too).
Brainstorm about ideas soon! It is also fine if it ispart of your own research.
Consult with me before starting
Best presentation award (for creditonly), after popular vote: a copy of thebook ³Vision Science´ (or another bookif the winner already has this one)
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 3/97
Class business
I have to travel
I am very flexible, but, how flexible are
you?
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 4/97
Wednesday
Sharing parts for intraclass transfer learning
Presenter: Sharat Chikkerur
Evaluator: Hueihan Jhuang
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 5/97
Some goals for this lecture
What are categories?
How many categories are there?
Some theories about the structure of object categories
Multiclass object recognition and transfer
learning in computer vision
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 6/97
An example of categorical perception
Continuous perception: graded response
5 0 1 00 1 50 20 0 25 0
50
10 0
15 0
20 0
25 0
Many perceptual phenomena are a mixture of the two: categorical at an everyday
level of magnification, but continuous at a more microscopic level. It can also
depend on cultural aspects, expertise, task, attention, «
5 0 10 0 15 0 2 0 0 2 5 0
50
10 0
15 0
20 0
25 0
Categorical perception: ³sharp´ boundaries
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 7/97
Another example
Continuous perception: graded response
20-24 25-29 30-34 35-39 40-44 45-49 50-54
Categorical perception: ³sharp´ boundaries
happinessfear
Emotions have categorical boundaries
Identification Task
% i d
e n t i f i c a t i o n
Anger
Fear
Happiness
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 8/97
Why do we care about categories?
Perception of function: We can perceive the 3D
shape, texture, material properties, withoutknowing about objects. But, the concept of
category encapsulates also information about
what can we do with those objects.
³We therefore include the perception of function as a proper ±indeed, crucial- subject
for vision science´, from Vision Science, chapter 9, Palmer .
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 9/97
Why do we care about categories?
When we recognize an object we can makepredictions about its behavior in the future,
beyond of what is immediately perceived.
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 10/97
The perception of function
Direct perception (affordances): GibsonFlat surface
Horizontal
Knee-high
«
Sittable
upon
One caveat of this comparison: deciding that
something is a chair might require access to more
features than the ones needed to decide that we can
sit on something« (it is a different level of
categorization)
Chair Chair
Chair?
Flat surface
Horizontal
Knee-high
«
Sittable
uponChair
Mediated perception (Categorization)
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 11/97
Direct perception
Some aspects of an object function can be
perceived directly
Functional form: Some forms clearly
indicate to a function (³sittable-upon´,
container, cutting device, «)
Sittable-upon Sittable-upon
Sittable-upon
It does not seem easy
to sit-upon this«
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 12/97
Direct perception
Some aspects of an object function can be
perceived directly
Observer relativity: Function is observer
dependent
From http://lastchancerescueflint.org
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 13/97
Limitations of Direct Perception
The functions are the same at some level of description: we can put things
inside in both and somebody will come later to empty them. However, weare not expected to put inside the same kinds of things«
Objects of similar structure might have very different functions
Not all functions seem to be available from direct visual information only.
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 14/97
Limitations of Direct Perception
Propulsion system
Strong protective surface
Something that looks like a door
Sure, I can travel to space onthis object
Visual appearance might be a very weak cue to function
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 15/97
Indirect perception of function by
categorization
Well« this requires object recognition (for
more details, see entire course)
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 16/97
So, what do we use direct or indirect?
³It seems exceedingly unlikely (though
logically possible) that we categorize
everything in our visual fields´, Palmer.
Hypothesis: we categorize the objects that
are relevant for a specific task that we
have at hand, but we only extract
affordances from the others.
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 17/97
How many categories?
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 18/97
Slide by Aude Oliva
³Muchas´
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 19/97
How many object categories are there?
Biederman 1987
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 20/97
How many categories?
Probably this question is not even specific
enough to have an answer
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 21/97
Which level of categorization
is the right one?Car is an object composed of:
a few doors, four wheels (not all visible at all times), a roof,
front lights, windshield
If you are thinking in buying a car, you might want to be a bit more specific about
your categorization.
?
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 22/97
psychological
cognition
content
belief
entity
object
artifact
structure
area
room
substance
Livingthing
organism
person
leader scientist
phenomenon
instrumentality
location
region
thing
animal
chordate
vertebrate
plant
chemist
body
stream
river
Categorical hierarchies
From Wordnet
Categories can be organized in hierarchies (tree structures are commonly used)
This is a mapping of the Wordnet tree into the 2D plane
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 23/97
Organizing ³things´ into categories
(1) Feature-based (2) prototype
Definition:disassembling aconcept into a set of
featural components
Each feature is anessential element of the category: ³ for athing to be an X, itmust have thatfeature. Otherwise it isnot an X´
Categories are formedon the basis of characteristics features,
which describe thetypical model of thecategory
Characteristics featuresare commonly present inexemplars of concepts,but they are not alwayspresent.
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 24/97
Prototypical Theory
According to theprototype view, an objectwill be classified as aninstance of a category if itis sufficiently similar to
the prototype.
Similarity ~ the number of features shared betweenan object and the
prototype (however,some features should beweighted more heavily asbeing more central to theprototype than are other
features).
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 25/97
Prototype or Sum of Exemplars ?
Prototype Model Exemplars Model
Category judgments are made
by comparing a new exemplar
to the prototype.
Category judgments are made
by comparing a new exemplar
to all the old exemplars of a category
or to the exemplar that is the most
appropriate
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 26/97
Levels of Categorization
The idea of prototypes and typicality led to the study of levels of categorization.
Rosch et al: ³albeit concepts exist at many different
levels of a hierarchy, one level is fundamental: basic
level.´
Basic level: the best compromise between grouping
together similar objects, and distinguish among
objects from the same category.
Willingham (247)
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 27/97
Levels of Categorization
B ASIC-LEVELCATEGORIES
SUPERORDIN ATE LEVELCATEGORIES
SUBORDIN ATE LEVELCATEGORIES
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 28/97
Rosch¶s Levels of CategorizationDefinition of Basic Level:
Similar shape: Basic level categories are the highest-level category for which their members have similar shapes.
Similar motor interactions: « for which people interact with its members
using similar motor sequences.
Common attributes: « there are a significant number of attributes in
common between pairs of members.
Sub Basic Superordinate
similarity
Similarity declines only slightly
going from subordinate tobasic level, and then drops
dramatically.
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 29/97
Rosch et al (1976) found that when people
are shown pictures of objects, they identify
objects at a basic level more quickly than
they identified objects at higher or lower
levels.
Objects appear to be recognized first at
their basic level, and only afterwards theyare classified in terms of higher or lowers
level categories
Levels of Categorization
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 30/97
Typicality effects
Typicality: how good or common an
item is a member of a given category.
The typical exemplar is like arepresentation of the average (or central tendency)
But, the representation of a categoryvary with experience, so does the³typical´ exemplar
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 31/97
Entry-level categories(Jolicoeur, Gluck, Kosslyn 1984)
Typical member of a basic-level categoryare categorized at the expected level
Atypical members tend to be classified at
a subordinate level.
A bird An ostrich
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 32/97
We do not need to recognize the exact
category
A new class can borrow information from
similar categories
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 33/97
So, where is computer vision?
Well«
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 34/97
The first problem« the data
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 35/97
Datasets
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 36/97
Datasets
Language 106 samples
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 37/97
Datasets
Language 106 samples
Character recognition
(MNIST)
104 samples
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 38/97
Datasets
Language 106 samples
Character recognition
(MNIST)
104 samples
Visual object recognition 10
3
samples
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 39/97
Early datasetsLena: a dataset in one picture
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 40/97
Coil
The Columbia Object Image Library (COIL-100) dataset consists of color images of 100objects where the images of the objects were taken at pose intervals of 5 degrees, i.e., 72
poses per object.
http://www.cs.columbia.edu/CA
VE
S. A. Nene, S. K. Nayar and H. Murase,
Technical Report CUCS-006-96, February 1996.
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 41/97
Collecting datasets
(towards 106-7 examples)
ESP game (CMU)Luis Von Ahn and Laura Dabbish 2004
LabelMe (MIT)Russell, Torralba, Freeman, 2005
StreetScenes (C
BC
L-MIT)Bileschi, Poggio, 2006
WhatWhere (Caltech)Perona et al, 2007
P ASCAL challenge
2006, 2007
Lotus Hill InstituteSong-Chun Zhu et al 2007
Cortina
UCSB
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 42/97
Labeling with games
L. von Ahn, L. Dabbish, 2004; L. von Ahn, R. Liu and M. Blum, 2006
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 43/97
The PASCAL Visual Object
Classes Challenge 2007
M. Everingham, Luc van Gool , C. Williams, J. Winn, A. Zisserman 2007
The twenty object classes that have been selected are:
Person: person
Animal: bird, cat, cow, dog, horse, sheep
Vehicle: aeroplane, bicycle, boat, bus, car, motorbike, train
Indoor: bottle, chair, dining table, potted plant, sofa, tv/monitor
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 44/97
Caltech 101 & 256
Griffin, Holub, Perona, 2007
Fei-Fei, Fergus, Perona, 2004
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 45/97
How to evaluate datasets?
How many labeled examples? How many classes? Segments or bounding
boxes? How many instances per image? How small are the targets? Variability
across instances of the same classes (viewpoint, style, illumination). How
different are the images?
How representative of the visual world is? What happens if you nail it?
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 46/97
Labelme.csail.mit.edu B. Russell, A. Torralba, K. Murphy, W.T. Freeman. IJCV 2008
Tool went online July 1st, 2005
290,000 object annotations
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 47/97
Polygon quality
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 48/97
« things do not always look good«
Polygon quality
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 49/97
Testing
Most common labels:
test
adksdsa
woiieiie
«
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 50/97
Online HooligansDo not try this at home
T lit
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 51/97
Tags quality
T lit
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 52/97
Tags quality
T lit
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 53/97
Tags quality
T lit
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 54/97
Tags quality
M tl b t lb
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 55/97
Matlab toolbox
LMquer y (database, 'object.name', 'car,building,road,tree')
St t
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 56/97
Stats
How many categories are in LabelMe?
How many levels there are?
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 57/97
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 58/97
10% of the objects
account for 90% of
the data
~Zipf¶s law
Caltech 101
Tiny images
LabelMe
We need transfer learning
Slid i +
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 59/97
Slide spain+perona
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 60/97
We have better low and mid-level vision
Better learning algorithms
Lot¶s of computational power
And lot¶s of data
«
We are running out of excuses
M lti l bj t d t ti
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 61/97
Multiclass object detectionthe not so early days
M lti l bj t d t ti
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 62/97
Multiclass object detectionthe not so early days
Schneiderman-Kanade multiclass object detection
Using a set of independent binary classifiers was a common strategy: Viola-Jones extension for dealing with rotations
- two cascades for each view
(a) One detector for each class
There is nothing wrong with this approach if you have access to
lots of training data and you do not care about efficiency.
Some symptoms of one vs all
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 63/97
Some symptoms of one-vs-all
multiclass approaches
Some of these parts cannot be used for anything else than this object.
What is the best representation to detect a traffic sign?
Very regular object: template matching will do the job
Parts derived from
training a binaryclassifier.
~100%
detection ratewith 0 false alarms
Some symptoms of one vs all
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 64/97
Some symptoms of one-vs-all
multiclass approachesPart-based object representation (looking for meaningful parts):
A. Agarwal and D. Roth
These studies try to recover parts that are meaningful. But is this the
right thing to do? The derived parts may be too specific, and they are
not likely to be useful in a general system.
M. Weber, M. Welling and P. Perona
Some symptoms of one vs all
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 65/97
Some symptoms of one-vs-all
multiclass approachesComputational cost grows linearly with Nclasses * Nviews * Nstyles «
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 66/97
Shared features
Is learning the object class 1000 easier than learning the first?
Can we transfer knowledge from one
object to another?
A
re the shared properties interesting bythemselves?
«
M ltit k l i
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 67/97
Multitask learning
R. Caruana. Multitask Learning. ML 1997
³MTL improves generalization by leveraging the domain-specific information
contained in the training signals of related tasks. It does this by training tasks in
parallel while using a shared representation´.
vs.
Sejnowski & Rosenberg 1986; Hinton 1986; Le Cun et al. 1989; Suddarth &
Kergosien 1990; Pratt et al. 1991; Sharkey & Sharkey 1992; «
M ltit k l i
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 68/97
Multitask learning
horizontal location of doorknob
single or double door
horizontal location of doorway center
width of doorway
horizontal location of left door jamb
horizontal location of right door jamb
width of left door jamb
width of right door jamb
horizontal location of left edge of door
horizontal location of right edge of door
Primar y task: detect door knobs
Tasks used:
R. Caruana. Multitask Learning. ML 1997
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 69/97
Sharing invariancesS. Thrun. Is Learning the n-th Thing Any Easier Than Learning The First?
NIPS 1996
Knowledge is transferred between tasks via a learned model of the
invariances of the domain: object recognition is invariant to rotation,
translation, scaling, lighting, « These invariances are common to all
object recognition tasks.
Toy world
Without sharing
With sharing
Convolutional Neural Network
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 70/97
Convolutional Neural Network
Translation invariance is already built into the network
The output neurons share all the intermediate levels
Le Cun et al, 98
Sharing transformations
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 71/97
Sharing transformations
Miller, E., Matsakis, N., and Viola, P. (2000). Learning from one example
through shared densities on transforms. In IEEE Computer Vision andPattern Recognition.
Transformations are sharedand can be learnt from other tasks.
M d l f bj t iti
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 72/97
Models of object recognitionI. Biederman, ³Recognition-by-components: A theory of human image
understanding,´ Psychological Review , 1987.
M. Riesenhuber and T. Poggio, ³Hierarchical models of object recognition in
cortex,´ Nature Neuroscience 1999.
T. Serre, L. Wolf and T. Poggio. ³Object recognition with features inspired
by visual cortex´. CVPR 2005
Sharing in constellation models
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 73/97
Sharing in constellation models(next Wednesday)
Pictorial StructuresF ischler & Elschlager, IEEE Trans. Comp. 1973
ConstellationModel
F ergus, Perona, & Zisserman, CVPR 2003
SVM DetectorsHeisele, Poggio, et. al., NIPS 2001
Model-Guided Segmentation
Mori, Ren, Efros, & Malik, CVPR 2004
R bl P t
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 74/97
Reusable Parts
Goal: Look for a vocabulary of edges that reduces the number of
features.
Krempp, Geman, & Amit ³Sequential Learning of Reusable Parts for Object
Detection´. TR 2002
N u m b e r o f f e a
t u r e s
Number of classes
Examples of reused parts
Additive models and boosting
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 75/97
Additive models and boosting(more details on Wednesday)
Torralba, Murphy, Freeman. CVPR 2004. P AMI 2007
Screen detector
Car detector
Face detector
Binary classifiers that share features:
Screen detector
Car detector
Face detector
Independent binary classifiers:
S ifi f t
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 76/97
Specific feature
Non-shared feature: this feature
is too specific to faces.
pedestrian
chair
Traffic light
sign
face
Background class
Sh d f t
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 77/97
Shared feature
shared feature
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 78/97
50 training samples/class
29 object classes
2000 entries in the dictionary
Results averaged on 20 runs
Error bars = 80% interval
Torralba, Murphy, Freeman. CVPR 2004. P AMI 2007
Shared features
Class-specific features
enera za on as a unc on o o ec
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 79/97
similarities
12 viewpoints12 unrelated object classes
Number of training samples per class Number of training samples per class
A r e a u n d e r R O C
A r e a u n d e r R O C K = 2.1 K = 4.8
Torralba, Murphy, Freeman. CVPR 2004. P AMI 2007
Sharing patches
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 80/97
Sharing patches
Bart and Ullman, 2004For a new class, use only features similar to features that where good for other
classes:
Proposed Dog
features
Transfer Learning for Image Classification with Sparse
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 81/97
Transfer Learning for Image Classification with Sparse
Prototype Representations
md d d
m
m
W W W
W W W
W W W
,2,1,
,22,21,2
,12,11,1
.
/11/
.
.
Coefficients for
for feature 2
Coefficients for
classifier 2
A. Quattoni, M. Collins, T. Darrell, CVPR 2008
§ §§! !
m
k
d
i
ik k
k
D y xk
W C y x f l D
k 1 1),(
|)(|max)),((||
1min
W
Hi hi l T i M d l
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 82/97
Hierarchical Topic Models
Topic models typically use a³bag of words´ approx.: ± Learning topics allows transfer
of information within a corpus of related documents
± Mixing proportions capture thedistinctive features of particular documents
T
U
z
x
J N
K
Latent Dirichlet Allocation (LDA)Blei, Ng, & Jordan, JMLR 2003
Pr(topic | doc)
Pr(word | topic)
E
Hi hi l T i M d l
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 83/97
Hierarchical Topic Models
T
U
z
x
J N
K
Latent Dirichlet Allocation (LDA)Blei, Ng, & Jordan, JMLR 2003
Pr(topic | doc)
Pr(word | topic)
E
Pr(x=word | z=topic) Pr(z=topic | doc)7topic
Pr(x=word | doc) =
=
Hierarchical Topic Models
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 84/97
Hierarchical Topic Models
T
U
z
x
J N
K
Latent Dirichlet Allocation (LDA)Blei, Ng, & Jordan, JMLR 2003
Pr(topic | doc)
Pr(word | topic)
E
³bag of features´ models:
Object Recognition ( Sivic et. al., ICCV 2005)
Scene Recognition ( F ei-F ei et. al., CVPR 2005)
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 85/97
Learning Shared Parts
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 86/97
Learning Shared Parts
Objects are often locally similar in appearance
Discover parts shared across categories
± How many total parts should we share?
± How many parts should each category use?
HDP Object Model
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 87/97
HDP Object Model
We learn the
number of parts.
Each object
uses a different
number of parts.
The model
assumes aknown number
of object
categories.
Parts are distributions
over appearances andlocations
HDP Object Model
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 88/97
HDP Object Model
There is no context, so the model is happy in creating
impossible part combinations.
Sharing Parts: 16 Categories
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 89/97
Sharing Parts: 16 Categories
Caltech 101 Dataset (Li & Perona)
Horses (Borenstein & Ullman)
C
at & dog faces (Vidal-Naquet & Ullman)
Bikes from Graz-02 (Opelt & Pinz)
Google«
Visualization of Shared Parts
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 90/97
Visualization of Shared Parts
Pr(appearance | part)
Pr(position | part)
Visualization of Shared Parts
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 91/97
Visualization of Shared Parts
Pr(appearance | part)
Pr(position | part)
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 92/97
Detection Results
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 93/97
6 Training Images per Categor y Detection vs. Training Set Size
Recognition Task
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 94/97
g
VS.
Recognition Results
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 95/97
g
6 Training Images per Categor y Recognition vs. Training Set Size
Recognition performance decreases. By sharing features, the classes look more similar.
Some more references
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 96/97
Some more references
Baxter 1996
Caruana 1997
Schapire, Singer, 2000
Thrun, Pratt 1997
Krempp, Geman, Amit, 2002
E.L.Miller, Matsakis, Viola, 2000
Mahamud, Hebert, Lafferty, 2001
Fink et al. 2003, 2004
LeCun, Huang, Bottou, 2004
Holub, Welling, Perona, 2005
«
Wednesday
8/3/2019 MIT6870_ORSU_lecture3: Levels of categorization and multiclass object recognition
http://slidepdf.com/reader/full/mit6870orsulecture3-levels-of-categorization-and-multiclass-object-recognition 97/97
y
Sharing parts for intraclass transfer learning
Presenter: SharatC
hikkerur
Evaluator: Hueihan Jhuang