Kind: captions
Language: en
sometimes we just don't have that friend
nearby that can tell us what a sign like
this might mean in a photo or a video
and it's not just about translating
languages
what if we wanted to know who this
person is on the right in this image or
be able to type this text that you can
see in the back of this conference hall
and run a google search on it
well thankfully there's a fix and it's
all available within a few clicks of
your mouse
so sit tight and get ready as we look at
how you can extract text from an image
to use it for searches translations and
furthering your research
[Music]
hi everyone and welcome back to this
series on how to do open source
investigations from home
i'm ben and this is part 12. so let's
get started
to start this session off let's say we
have this photo of someone holding up
this sign at a protest
now we could ask a friend what it means
sure
and i'd always recommend you get a
second opinion on translations rather
than using machines for translations
because well as we know machines aren't
always right but i've got an english
keyboard in front of me and i can't
really type out what that sign says if i
wanted to search for it in google so for
a sign like this maybe i can use
something like google translate's draw
function
and start to really write out some of
these characters but we could actually
do this a little bit easier and there's
something that's really quite handy that
exists called ocr
ocr stands for optical character
recognition
it allows the text to be pulled out of
an image so we can either use that in
translations or we can use that to run a
search in google or another search
engine platform as well
one of my favorite apps to do this is
yandex and this is really helpful if
you're doing this from an image reverse
search perspective so say you wanted to
reverse that search that image that you
found in an article and you want to get
a translation out of that or pull that
text out of it all we need to do is
right click or upload that using this
icon that you see in the middle there
and already you can see on the right of
this image that we have the text that's
been pulled out of that image i can
easily just click on that translate
function
and that'll bring that up in translate
and what we've got is one for all and
all for one which is what that banner
says on that image right there so what
about other uses for this because
perhaps this might be simple for some
people perhaps it might be written in an
article already or wherever the context
is of the image that's been used so for
example what if we wanted to identify
this person on the right well we've got
a little name written there don't we so
if i was to right click on this and i
was to run an image reverse search on
this in yandex perhaps we can click on
recognize text
now there's a couple of bits of text
that we can see in there it seems like
this one has picked this up and it's
always nice to just double check to make
sure that we've got that
it's picked up getty images as well and
it's also picked up that name
we could use the google translate
function on that but what we can also do
which is really useful about pulling
that text out of the image is that we
can just copy that text
and run it in a search engine
and what i'm doing now is i'm running a
google search
for the text that we've just pulled out
of that image of that name plate right
there which is the original spelling of
that person's name and we can see that
that's the vice president of the
people's republic of china cool so what
about another example well how about we
look at what this actual phrase says on
the back of this image here
uh from a conference hall in north korea
well again we can take it through the
same process
i always like to try and get the biggest
and most clear image possible to reverse
search that or to run that through any
platform that you might have your hands
on
and simply run that through yandex as
well if you don't have any plugins that
allow you to right click and go to
yandex you can simply save the image and
just upload as an image of our search
into the index so you can see that
that's pulled out that text for us and
already we can just select that and with
my google translate plugin we can see
what that might say
now i always like to just double check
with all these characters as well
because sometimes this might just be off
or it might be a little bit inaccurate
also sometimes the
translation function on google isn't
really that great for so many different
languages but one thing we can do is
just run a simple search on that text as
well and see perhaps if that comes up
with any different results yes we've got
a few different results which is quite
interesting so you're probably thinking
okay this seems pretty easy for small
bits of text but it's also for big bits
of text as well and it's not just on
yandex there's lots of ocr bits of
software out there and platforms i'd
recommend you vet each one of them
before you start uploading any private
documents or private images that are not
on the internet
but however here's one
called i2 ocr i really like using this
one because you have two different
options so you can do image ocr or you
can do pdf ocr
the reason why pdf ocr on this one is
quite good in comparison to
platforms like google is because this
one offers very different languages so
for example we could run a pdf
in burmese and we could get that loosely
translated again i recommend you get
someone that speaks burmese or can read
burmese to actually confirm any
translations before you rely on them in
any publications or anything like that
so in testing out how this platform
actually operates let's try an example
such as this which is a post from
december 25 from the sudan doctors
committee now often you'll see
perhaps facebook posts or you might not
know that it's from facebook but it
might be posted as a screenshot on
twitter
and because it's not raw text but rather
a screenshot of the post you might not
know or might not be able to easily
translate this
into google or something like that so a
really easy method to do that is just to
take that screenshot and really run that
through one of these ocr platforms so
we'll do that now
i've selected my input language and i
can select the file here we definitely
want to make sure that we are not a
robot
so sometimes it's always fun to do a
little bit of spot detractors and we'll
extract the text so what we have here is
the raw text on the left that i can
select and pop into a translate function
or something like that
on the right we can see how that has
been picked up
from that image that we originally
uploaded that screenshot image so the i2
ocr site is quite useful because we can
just take that text and we can run that
in either bing or google
and actually translate that text so you
can see that it's quite useful to
translate lots of text but of course
this is clearly written text that you
can see from that screenshot there if
for example you were to have a street
sign or a sign of a shop that you can
see in a video but isn't really
easily replicable for something like
yandex or something like this where we
can we not able to pull out that text
i'd recommend going back to the draw
function and just having a little go at
drawing that text in there whether it's
arabic
or whether it's cyrillic or some other
form of alphabet that you just can't
clearly get that ocr working on so there
we have it i hope you found this session
on extracting text from images helpful
and useful
if this type of technique has helped you
in your research or does help you please
do let me know in the comments section
below
i also just wanted to say a big thank
you for all of the comments likes and
subscribes to this channel
especially the messages from journalists
researchers and investigators in
developing countries who are learning
this content and using it to hold those
responsible to account
i'm so glad i can be of such help to you
all
i'll see you in the next session soon
until then take it easy
[Music]
you