ICCV Daily - Friday

12 / 24

The work tackles domain adaptation,
which is an interesting and challenging
problem in computer vision. There are
already many works on the subject of
domain adaptation and recently on
deep domain adaptation – that is how
to handle domain shift in deep
networks. Domain shift means that
there are huge discrepancies between
your training data and your test
setting.
The main advantage of the team’s
approach is simplicity. They have layers
called domain alignment layers that
can be embedded in any architecture
to cope with this domain shift problem
and reduce it. They have a method
that can be implemented in a few lines
of code which outperforms the state of
the art in one of the most challenging
problems in the community.
“ To my knowled
other benchm
12
Friday
Elisa Ricci
AutoDIAL: Automatic DomaIn Alignment Layers
Elisa Ricci is Assistant Professor at the Department of Engineering at
the University of Perugia and a researcher in the Technologies of Vision
group at Fondazione Bruno Kessler.
Elisa will present a poster co-authored with Fabio Maria Carlucci,
Lorenzo Porzi, Barbara Caputo, and Samuel Rota Bulò. It is joint work
between Fondazione Bruno Kessler, and La
Sapienza University of Rome.
From left: