<s>
It	O
uses	O
a	O
Grasp	O
Quality	O
Convolutional	B-Architecture
Neural	I-Architecture
Network	I-Architecture
to	O
learn	O
how	O
to	O
grasp	O
unusually	O
shaped	O
objects	O
.	O
</s>
<s>
Dex-net	O
includes	O
a	O
high-resolution	O
3-D	O
sensor	O
and	O
two	O
arms	O
,	O
each	O
controlled	O
by	O
a	O
different	O
neural	B-Architecture
network	I-Architecture
.	O
</s>
<s>
The	O
robot	O
’s	O
software	O
scans	O
an	O
object	O
and	O
then	O
asks	O
both	O
neural	B-Architecture
networks	I-Architecture
to	O
decide	O
,	O
on	O
the	O
fly	O
,	O
whether	O
to	O
grab	O
or	O
suck	O
a	O
particular	O
object	O
.	O
</s>
<s>
In	O
a	O
contest	O
organized	O
by	O
Amazon	B-Application
recently	O
,	O
the	O
best	O
robots	O
were	O
capable	O
of	O
between	O
70	O
and	O
95	O
.	O
</s>
