<s>
Caltech	B-General_Concept
101	I-General_Concept
is	O
a	O
data	B-General_Concept
set	I-General_Concept
of	O
digital	B-Algorithm
images	I-Algorithm
created	O
in	O
September	O
2003	O
and	O
compiled	O
by	O
Fei-Fei	O
Li	O
,	O
Marco	O
Andreetto	O
,	O
Marc	O
'	O
Aurelio	O
Ranzato	O
and	O
Pietro	O
Perona	O
at	O
the	O
California	O
Institute	O
of	O
Technology	O
.	O
</s>
<s>
It	O
is	O
intended	O
to	O
facilitate	O
Computer	B-Application
Vision	I-Application
research	O
and	O
techniques	O
and	O
is	O
most	O
applicable	O
to	O
techniques	O
involving	O
image	O
recognition	O
classification	O
and	O
categorization	O
.	O
</s>
<s>
Caltech	B-General_Concept
101	I-General_Concept
contains	O
a	O
total	O
of	O
9,146	O
images	O
,	O
split	O
between	O
101	O
distinct	O
object	O
categories	O
(	O
faces	O
,	O
watches	O
,	O
ants	B-Application
,	O
pianos	O
,	O
etc	O
.	O
)	O
</s>
<s>
Provided	O
with	O
the	O
images	O
are	O
a	O
set	O
of	O
annotations	O
describing	O
the	O
outlines	O
of	O
each	O
image	O
,	O
along	O
with	O
a	O
Matlab	B-Language
script	B-Language
for	O
viewing	O
.	O
</s>
<s>
Most	O
Computer	B-Application
Vision	I-Application
and	O
Machine	O
Learning	O
algorithms	O
function	O
by	O
training	O
on	O
example	O
inputs	O
.	O
</s>
<s>
For	O
example	O
,	O
the	O
real-time	O
face	B-General_Concept
detection	I-General_Concept
method	O
used	O
by	O
Paul	O
Viola	O
and	O
Michael	O
J	O
.	O
Jones	O
was	O
trained	O
on	O
4,916	O
hand-labeled	O
faces	O
.	O
</s>
<s>
Historically	O
,	O
most	O
data	B-General_Concept
sets	I-General_Concept
used	O
in	O
computer	B-Application
vision	I-Application
research	O
have	O
been	O
tailored	O
to	O
the	O
specific	O
needs	O
of	O
the	O
project	O
being	O
worked	O
on	O
.	O
</s>
<s>
A	O
large	O
problem	O
in	O
comparing	O
computer	B-Application
vision	I-Application
techniques	O
is	O
the	O
fact	O
that	O
most	O
groups	O
use	O
their	O
own	O
data	B-General_Concept
sets	I-General_Concept
.	O
</s>
<s>
The	O
Caltech	B-General_Concept
101	I-General_Concept
data	B-General_Concept
set	I-General_Concept
aims	O
at	O
alleviating	O
many	O
of	O
these	O
common	O
problems	O
.	O
</s>
<s>
Available	O
for	O
general	O
use	O
,	O
Caltech	B-General_Concept
101	I-General_Concept
acts	O
as	O
a	O
common	O
standard	O
by	O
which	O
to	O
compare	O
different	O
algorithms	O
without	O
bias	O
due	O
to	O
different	O
data	B-General_Concept
sets	I-General_Concept
.	O
</s>
<s>
However	O
,	O
a	O
recent	O
study	O
demonstrates	O
that	O
tests	O
based	O
on	O
uncontrolled	O
natural	O
images	O
(	O
like	O
the	O
Caltech	B-General_Concept
101	I-General_Concept
data	B-General_Concept
set	I-General_Concept
)	O
can	O
be	O
seriously	O
misleading	O
,	O
potentially	O
guiding	O
progress	O
in	O
the	O
wrong	O
direction	O
.	O
</s>
<s>
The	O
Caltech	B-General_Concept
101	I-General_Concept
data	B-General_Concept
set	I-General_Concept
consists	O
of	O
a	O
total	O
of	O
9,146	O
images	O
,	O
split	O
between	O
101	O
different	O
object	O
categories	O
,	O
as	O
well	O
as	O
an	O
additional	O
background/clutter	O
category	O
.	O
</s>
<s>
A	O
Matlab	B-Language
script	B-Language
is	O
provided	O
with	O
the	O
annotations	O
.	O
</s>
<s>
It	O
loads	O
an	O
image	O
and	O
its	O
corresponding	O
annotation	O
file	O
and	O
displays	O
them	O
as	O
a	O
Matlab	B-Language
figure	O
.	O
</s>
<s>
The	O
Caltech	B-General_Concept
101	I-General_Concept
data	B-General_Concept
set	I-General_Concept
was	O
used	O
to	O
train	O
and	O
test	O
several	O
computer	B-Application
vision	I-Application
recognition	O
and	O
classification	O
algorithms	O
.	O
</s>
<s>
The	O
first	O
paper	O
to	O
use	O
Caltech	B-General_Concept
101	I-General_Concept
was	O
an	O
incremental	O
Bayesian	O
approach	O
to	O
one	O
shot	O
learning	O
,	O
an	O
attempt	O
to	O
classify	O
an	O
object	O
using	O
only	O
a	O
few	O
examples	O
,	O
by	O
building	O
on	O
prior	O
knowledge	O
of	O
other	O
classes	O
.	O
</s>
<s>
The	O
Caltech	B-General_Concept
101	I-General_Concept
images	O
,	O
along	O
with	O
the	O
annotations	O
,	O
were	O
used	O
for	O
another	O
one	O
shot	O
learning	O
paper	O
at	O
Caltech	O
.	O
</s>
<s>
Other	O
Computer	B-Application
Vision	I-Application
papers	O
that	O
report	O
using	O
the	O
Caltech	B-General_Concept
101	I-General_Concept
data	B-General_Concept
set	I-General_Concept
include	O
:	O
</s>
<s>
Proceedings	O
of	O
2005	O
IEEE	O
Computer	O
Society	O
Conference	O
on	O
Computer	B-Application
Vision	I-Application
and	O
Pattern	O
Recognition	O
(	O
CVPR	O
2005	O
)	O
,	O
IEEE	O
Computer	O
Society	O
Press	O
,	O
San	O
Diego	O
,	O
June	O
2005	O
.	O
</s>
<s>
Caltech	B-General_Concept
101	I-General_Concept
has	O
several	O
advantages	O
over	O
other	O
similar	O
data	B-General_Concept
sets	I-General_Concept
:	O
</s>
<s>
Caltech	B-General_Concept
101	I-General_Concept
users	O
generally	O
do	O
not	O
need	O
to	O
crop	O
or	O
scale	O
images	O
before	O
they	O
can	O
be	O
used	O
.	O
</s>
<s>
Weaknesses	O
to	O
the	O
Caltech	B-General_Concept
101	I-General_Concept
data	B-General_Concept
set	I-General_Concept
may	O
be	O
conscious	O
trade-offs	O
,	O
but	O
others	O
are	O
limitations	O
of	O
the	O
data	B-General_Concept
set	I-General_Concept
.	O
</s>
<s>
Papers	O
that	O
rely	O
solely	O
on	O
Caltech	B-General_Concept
101	I-General_Concept
are	O
frequently	O
rejected	O
.	O
</s>
<s>
The	O
data	B-General_Concept
set	I-General_Concept
is	O
too	O
clean	O
:	O
</s>
<s>
The	O
Caltech	B-General_Concept
101	I-General_Concept
data	B-General_Concept
set	I-General_Concept
represents	O
only	O
a	O
small	O
fraction	O
of	O
possible	O
object	O
categories	O
.	O
</s>
<s>
Aliasing	B-Error_Name
and	O
artifacts	B-Algorithm
due	O
to	O
manipulation	O
:	O
</s>
<s>
Some	O
images	O
have	O
been	O
rotated	O
and	O
scaled	O
from	O
their	O
original	O
orientation	O
,	O
and	O
suffer	O
from	O
some	O
amount	O
of	O
artifacts	B-Algorithm
or	O
aliasing	B-Error_Name
.	O
</s>
<s>
Caltech	O
256	O
is	O
another	O
image	O
data	B-General_Concept
set	I-General_Concept
,	O
created	O
in	O
2007	O
.	O
</s>
<s>
It	O
is	O
a	O
successor	O
to	O
Caltech	B-General_Concept
101	I-General_Concept
.	O
</s>
<s>
It	O
is	O
intended	O
to	O
address	O
some	O
of	O
the	O
weaknesses	O
of	O
Caltech	B-General_Concept
101	I-General_Concept
.	O
</s>
<s>
Overall	O
,	O
it	O
is	O
a	O
more	O
difficult	O
data	B-General_Concept
set	I-General_Concept
than	O
Caltech	B-General_Concept
101	I-General_Concept
,	O
but	O
it	O
suffers	O
from	O
comparable	O
problems	O
.	O
</s>
<s>
LabelMe	B-General_Concept
is	O
an	O
open	O
,	O
dynamic	O
data	B-General_Concept
set	I-General_Concept
created	O
at	O
MIT	O
Computer	O
Science	O
and	O
Artificial	O
Intelligence	O
Laboratory	O
(	O
CSAIL	O
)	O
.	O
</s>
<s>
LabelMe	B-General_Concept
takes	O
a	O
different	O
approach	O
to	O
the	O
problem	O
of	O
creating	O
a	O
large	O
image	O
data	B-General_Concept
set	I-General_Concept
,	O
with	O
different	O
trade-offs	O
.	O
</s>
<s>
Users	O
may	O
add	O
images	O
to	O
the	O
data	B-General_Concept
set	I-General_Concept
by	O
upload	O
,	O
and	O
add	O
labels	O
or	O
annotations	O
to	O
existing	O
images	O
.	O
</s>
<s>
Due	O
to	O
its	O
open	O
nature	O
,	O
LabelMe	B-General_Concept
has	O
many	O
more	O
images	O
covering	O
a	O
much	O
wider	O
scope	O
than	O
Caltech	B-General_Concept
101	I-General_Concept
.	O
</s>
<s>
Overhead	B-General_Concept
Imagery	I-General_Concept
Research	I-General_Concept
Data	I-General_Concept
Set	I-General_Concept
(	O
OIRDS	O
)	O
is	O
an	O
annotated	O
library	O
of	O
imagery	O
and	O
tools	O
.	O
</s>
<s>
MICC-Flickr	O
101	O
is	O
an	O
image	O
data	B-General_Concept
set	I-General_Concept
created	O
at	O
the	O
Media	O
Integration	O
and	O
Communication	O
Center	O
(	O
MICC	O
)	O
,	O
University	O
of	O
Florence	O
,	O
in	O
2012	O
.	O
</s>
<s>
It	O
is	O
based	O
on	O
Caltech	B-General_Concept
101	I-General_Concept
and	O
is	O
collected	O
from	O
Flickr	B-Algorithm
.	O
</s>
<s>
MICC-Flickr	O
101	O
corrects	O
the	O
main	O
drawback	O
of	O
Caltech	B-General_Concept
101	I-General_Concept
,	O
i.e.	O
</s>
<s>
It	O
builds	O
on	O
a	O
standard	O
and	O
widely	O
used	O
data	B-General_Concept
set	I-General_Concept
composed	O
of	O
a	O
manageable	O
number	O
of	O
categories	O
(	O
101	O
)	O
and	O
therefore	O
can	O
be	O
used	O
to	O
compare	O
object	O
categorization	O
performance	O
in	O
a	O
constrained	O
scenario	O
(	O
Caltech	B-General_Concept
101	I-General_Concept
)	O
and	O
object	O
categorization	O
"	O
in	O
the	O
wild	O
"	O
(	O
MICC-Flickr	O
101	O
)	O
on	O
the	O
same	O
101	O
categories	O
.	O
</s>
