28 November 2020 | New Scientist | 13MindYOU can now estimate how many
calories are in a meal simply by
taking a photo of it.
Calorie counting is one of the
ways many people try to control
their weight, but manually entering
nutritional information about
ingredients into apps is time-
consuming. Cooking meals muddles
things further, making it difficult
to get accurate calorie counts.
Robin Ruede and his colleagues
at the Karlsruhe Institute of
Technology, Germany, might be
able to help. They have harnesseda commonly used neural network
called DenseNet to cross-reference
images of meals with a database
of 308,000 photographs taken
from 70,000 recipes on a German
cooking website. A neural network
is a software system modelled
on the architecture of a brain.
“We adapted the architecture
and made it predict the
macronutrients – such as
fat and protein content – from
the ingredients,” says Ruede.
“We assume they cooked the recipe
correctly, take the nutritional values
and make the model learn the
correlation between the nutritional
information and that image.”
The model is far from perfect:
on average, its estimate of caloriesCHOOSING between going out
for a run or staying slumped on
your sofa in front of the TV can
be tricky, but it turns out your
decision can be seen in your eyes
before you have even made it.
When we do something that
requires physical effort, our pupils
can dilate and activity heightens
in the prefrontal cortex, part of
the brain that is vital to cognition.
Now, it seems that these two
reactions may also guide our
decisions about activities that
we are thinking about doing.
To investigate this idea,
Irma Kurniawan and her
colleagues at the University
of Zurich in Switzerland asked
49 people to choose between
different tasks that varied by
the level of effort involved.
The researchers first got the
participants to do hand squeezes
using a handheld device at varying
degrees of physical difficulty.
Each person was then placed
inside a functional MRI scanner
to record their brain activitywhile an eye-tracker also
monitored their pupil size.
While in the scanner,
participants were asked to
choose between doing more
strenuous or effortless hand
contractions later on, with a
greater cash reward for choosing
the more difficult exercises. Onceoutside, 30 minutes to an hour
later, they completed a random
selection of hand squeezes at
their chosen levels of effort.
The team saw changes in
pupil size and prefrontal cortex
activity as people made their
decision in the scanner. Because
these changes occurred before
doing the exercises, it suggests
that the participants were
anticipating the amount of
effort that would be required.
What’s more, if someone chose
the most difficult activity, this was
revealed by specific pupil dilation
and brain activity patterns. The
team suggests that these signals
influence the outcome of people’s
decisions, by helping to predict
the amount of energy required,
and this reveals whether they will
end up doing a higher effort task
(bioRxiv, doi.org/fjp6).
“It’s a very interesting proposal,”
says Tobias Hauser at University
College London. However, pupil
size and the prefrontal cortex
signal are known to reflectdifferent things, he says.
“[They] have been linked to
different aspects of cognition,
be it effort, be it surprise, be it
difficulty, so it’s not a unitary
thing,” says Hauser.
As to whether pupil dilation
and prefrontal cortex activity
really play a role in our decision-
making about future exertion,“it’s a long shot, but long shots
are worth pursuing”, says Hauser.
Follow-up studies would
need to establish whether these
two factors directly affect our
behaviour and if they might act
differently in people with low
motivation, he says. If that holds
true, Hauser thinks “it’s definitely
an interesting new perspective on
effort and decision-making which
could in part reformulate what
we understand”. ❚Computing
Whether you intend to
exercise may be shown
in your eyesGege LiUT
A^ LAUTE
RB
AC
H/EYE
EM/GETT
Y^ IMAGESThe eyes have it
Your pupils can help reveal your decisions before you are conscious of making them
Software estimates
calorie content of
food from an imageChris Stokel-Walkerloaf of bread at 229kcal (arxiv.org/
abs/2011.01082).
“The whole paper is a big step
forward in our ability to determine
the nutritional value of food
from pictures,” says Dane Bell,
co-founder of Lum AI, a natural
language processing company.
“This data set directly bears on
what we want to know: how much
protein, carbs and fat this food has.”
The model falls down when
confronted with items that aren’t
in the list of recipes or when recipes
use unusual ingredients or methods.
But even so, says Ruede, “it’s pretty
clear it can distinguish between
categories of high-calorie and
low-calorie foods”. ❚One photograph
is now enough to
estimate a meal’s
calorie contentFILAD
ENDRON/GETTY^ IMAGESis 32.6 per cent awry when
confronted with a previously
unseen image, though humans
are also poor at estimating calorific
content: a 2018 survey found
our estimates can be hundreds
of calories out. In contrast, the
neural network model estimated
a chocolate cake, which was
198kcal per 100 grams, as being
183kcal, and a 239kcal/100g“ Changes in pupil size
seemed to anticipate
the amount of effort
exercises would require”