How Does Optical Character Recognition (OCR) Work?

  Рет қаралды 429,772

Techquickie

7 жыл бұрын

How do computers read text on a page, and how has the technology improved?
Freshbooks message: Head over to freshbooks.com/techquickie and don’t forget to enter Tech Quickie in the “How Did You Hear About Us” section when signing up for your free trial.
Techquickie Merch Store: www.designbyhumans.com/shop/LinusTechTips/
Techquickie Movie Poster: shop.crowdmade.com/collections/linustechtips/products/tech-quickie-24x36-poster
Follow: linustech
Join the community: linustechtips.com

Пікірлер: 419
@TheOriginalFayari
@TheOriginalFayari 7 жыл бұрын
That was the smoothest transition to a sponsor spot I've ever seen.
@thepalettewhispererasmr1227
@thepalettewhispererasmr1227 3 жыл бұрын
I didnt even realize it was happening
@jamesklein4399
@jamesklein4399 7 жыл бұрын
FILE FORMATS AS FAST AS POSSIBLE! png vs jpg mp4 vs mkv mp3 vs ...?
@laser5317
@laser5317 7 жыл бұрын
James Klein MP3 vs WAV
@RobertHildebrandt
@RobertHildebrandt 7 жыл бұрын
mp3 vs flac
@coffeen8128
@coffeen8128 7 жыл бұрын
James Klein png keep the quility
@smarthd7749
@smarthd7749 7 жыл бұрын
MP4 and .mkv Is not a file format, IT is a container. And ITS not many difference between mkv and MP4 the only difference is that mkv can hold some more codecs.
@cldream
@cldream 7 жыл бұрын
SmartFyrHD Also Matroska can also embed multiple subtitle formats (SRT, SSA/Advanced SSA)
@DisbelieverH2o
@DisbelieverH2o 7 жыл бұрын
I gotta say, I really liked this one! Very informative but what really made it for me was the seamless sponsor spot. I'd love to see more in such a way!
@freedomofmotion
@freedomofmotion 7 жыл бұрын
Irish travelers will be deeply hurt that OCR and even you don't accept that dag is a word. Has no one ever tried to sell you a dag? Or admired your dag?
@chantafreak
@chantafreak 7 жыл бұрын
Ya like dags?
@ataksnajpera
@ataksnajpera 7 жыл бұрын
Knackers do not even speak english ;)
@GewelReal
@GewelReal 7 жыл бұрын
hey kid, you wanna buy some dags?
@EvadingFate
@EvadingFate 7 жыл бұрын
Oh, dogs. Sure, I like dags. I like caravans more.
@chantafreak
@chantafreak 7 жыл бұрын
This is the post I was waiting for.
@DustinRodriguez1_0
@DustinRodriguez1_0 7 жыл бұрын
OCR was one of the first practical uses of neural networks back in the 70s or 80s. Maybe even earlier? When I took an AI class in college, we wrote a simple OCR neural net and it was pretty easy.
@jandresshade
@jandresshade 7 жыл бұрын
the OCR can use different techniques to recognize character, one is creating a model based on data of different characters and training the sofware to recognize them( Artificial neural networks is an example of this)
@ziyitan8996
@ziyitan8996 7 жыл бұрын
I love how Luke explains stuff :D
@SnypeSin
@SnypeSin 7 жыл бұрын
that's good and all but I would have thought you'd give us and idea of what kind of devices use OCR for consumer/business.
@Mr.FastZombie
@Mr.FastZombie 7 жыл бұрын
There are also programs for character recognition on your screen. Project Naptha is a Chrome extension that can let you copy and paste words in an image. And ShareX has OCR that you can use for any program.
@OMNIA_RH
@OMNIA_RH 5 жыл бұрын
Thank so much for you explaining Sir.
@ShreyPandya150
@ShreyPandya150 7 жыл бұрын
When Luke said it wouldn't look as crisp and the video resolution went down I instantly checked if I was at 1080p
@soroushjm1011
@soroushjm1011 4 жыл бұрын
Yeah me too
@sabaamin3179
@sabaamin3179 2 жыл бұрын
Just what I was looking for. Good Job!
@TheDyingFox
@TheDyingFox 7 жыл бұрын
I was going to ask "How about Voice Recognition next?" but searched your channel, and I'll be damned, 1 year ago, you guys work fast! (Not sure how I've been missing it though, alot of content much?). It's a shame neither is "How to create your own Voice Recognition and Optical Character Recognition as fast as possible"
@HirooKoslov
@HirooKoslov 7 жыл бұрын
My ScanSnap IX500 usese software to make scans readable. It works pretty well and the IX500 is blisteringly fast.
@quenjankosky7348
@quenjankosky7348 7 жыл бұрын
Well, with OCR, there is an exception for the lack of accuracy. When basic modern OCR was being developed, they made a series of fonts deigned to be as accurate as possible. These fonts were OCR-A and OCR-B. These fonts are super accurate with OCR, and there is usually never any error with them.
@pearls9133
@pearls9133 7 жыл бұрын
could you do videos explaining how mastering audio and video works? (if it doesnt already exist)
@cestsibon2468
@cestsibon2468 3 жыл бұрын
This is the first time i've watched a tech video and actually not had a headache after. Waiting for the interpretive google dance hehe
@narutosasuke30
@narutosasuke30 5 жыл бұрын
Which OCR recognizes Handwritten text that you have shown at the end? I couldn't find anything which actually does that within a permissible error rate :/
@arnatsemtappra3822
@arnatsemtappra3822 6 жыл бұрын
Very useful knowledge and easy to understand provided to the new faces of this technology.
@macpclinux1
@macpclinux1 7 жыл бұрын
luke are you finally using linux? i saw that little ubuntu font box :D good job mate!
@ulashofficial
@ulashofficial 4 жыл бұрын
Sir can you tell me how can i find duplicate numbers with any OCR app or how should i pursue to make an app for that ?
@JRDev4All
@JRDev4All 7 жыл бұрын
You should do an as fast as possible on assistive technologies such as screen readers
@dav2mai
@dav2mai 7 жыл бұрын
Will it also recognize language? because "dag" translates to "day" in Danish
@Meg_A_Byte
@Meg_A_Byte 7 жыл бұрын
Is there anything on this world that recognizes danish?
@22RH544
@22RH544 7 жыл бұрын
Nope, as a Dutch guy i can read it just fine, but when it is spoken.................I quit.
@TheDyingFox
@TheDyingFox 7 жыл бұрын
Same result when translated to Swedish xD
@Mr.FastZombie
@Mr.FastZombie 7 жыл бұрын
I would assume it sticks to one language, but some can probably change their language. Also perhaps some could be able to determine the language based on what it has already recognized.
@crewskater06
@crewskater06 7 жыл бұрын
It's from the movie Snatch
@moenbase1
@moenbase1 2 жыл бұрын
In my industry, which is electronics. We use OCR in our automated optical machine to detect component marking on components as small as micro BGA's that are like 400microns wide. It's amazing to see how you can push it's limits. Just, sometimes like when there's a sufficient amount of flux on the components it makes it impossible to read.
@Ahmed71616
@Ahmed71616 2 жыл бұрын
What is the best scanner that does the same job as your devices
@HolarMusic
@HolarMusic 7 жыл бұрын
Is that an 8k green-screen video? Looks super clean
@hillppari
@hillppari 7 жыл бұрын
Google translate app with OCR is pretty nifty when you can translate foreign signs etc.
@KX36
@KX36 7 жыл бұрын
I did some OCR recently. Tesseract on Linux was the best at recognising the text accurately, but it outputs plain text only. There are 3rd party GUIs, but still none really preserve formatting. ABBYY FineReader on Windows (the gold standard for home use) was quite good at preserving formatting but worse at recognising text accurately. My scan was 200 pages of black 12pt Times New Roman on white paper scanned at 300dpi which should be one of the easiest things to process, and it regularly made mistakes on 1 vs l vs I , y vs v, H vs II etc. And these were often in places the dictionary should have easily known what it should have been. How often do you get a lower case L in the middle of a long number or a double upper case I at the start of a word or a v at the end of a word. It took 3 hours to go through the document correcting the mistakes it highlighted. Don't know how many mistakes are in there that it didn't highlight.
@littletomatomonkeysmeeeeel8324
@littletomatomonkeysmeeeeel8324 Жыл бұрын
Highly recommend PaddleOCR! 80 languages supported! Good performance! Easy to use! It would be great if bloggers could do a comparative evaluation of the popular OCR tools.
@94213915
@94213915 5 жыл бұрын
Can you please tell me about any OCR software for devanagari language . Which can cost me less
@JOELwindows7
@JOELwindows7 7 жыл бұрын
Wow, I saw this video right near before my National exam days.
@howardt12345
@howardt12345 7 жыл бұрын
Dennis: "You are dancing?"
@MiMiOrt
@MiMiOrt 3 жыл бұрын
I downloaded but , I thought that it will recognize the different fonts that are someonetimes in just ONE page. Does anyone know an APP/Program that can recognize the font on a scanned document?
@rushabmehta
@rushabmehta 7 жыл бұрын
Can you do video on Virtualization such as hardware, network and storage Virtualization.
@TheZorch
@TheZorch 7 жыл бұрын
I've got a Chrome extension that does OCR within images. Sometimes comes in really handy.
@vapexxx
@vapexxx 7 жыл бұрын
Luke - I actually watched the ad because of your fresh moves!
@Lorten369
@Lorten369 7 жыл бұрын
YEES More history please. love knowledge.
@bradad1111
@bradad1111 7 жыл бұрын
Saw OCR and immediately thought it had something to do with the Exam Board.
@craigmalcom6294
@craigmalcom6294 7 жыл бұрын
bradad111 Lool same
@StickyBagel
@StickyBagel 5 жыл бұрын
So did youtube, i was watching a revision playlist and here i am??
@fleksimir
@fleksimir 4 жыл бұрын
Linus ad (pulseway) on linus video. I love this ahahaha
@jankomirovic2866
@jankomirovic2866 3 жыл бұрын
same gahahahahah
@leivadaros
@leivadaros 7 жыл бұрын
Haven't read a single comment regarding the video's topic.... only "First", "Notification Squad where you at" and comments trying to be witty..... Great video by the way, i love getting general introductory information on the subject of my studies (computer engineer). Keep at it TechQuickie :D
@rry1994
@rry1994 7 жыл бұрын
I love u guys man
@jehdo144
@jehdo144 7 жыл бұрын
great video!
@jamilangon5798
@jamilangon5798 7 жыл бұрын
well google releases a OCRT (optical character recognition translator). which translate even other character aside from ASCII (chinese, japanese, thai and other non alpha character)... it become useful for those who travel and find themselves trap into a place where no one can speak or understand english.
@sebon11
@sebon11 4 жыл бұрын
Cool! Thx a lot.
@antonjohansson1384
@antonjohansson1384 7 жыл бұрын
Dag is in swedish day
@jean-lucasymptotic5083
@jean-lucasymptotic5083 7 жыл бұрын
Speaking of machine learning..... that would make a good techquickie :D
@MotivationAdonis
@MotivationAdonis 7 жыл бұрын
Linus tech tips as fast as possible
@Jinni_SD
@Jinni_SD 7 жыл бұрын
I really like Tesseract withHomebrew on Mac for OCR.
@teksight9714
@teksight9714 7 жыл бұрын
Good video. Thumbs up!
@rediculousman
@rediculousman 7 жыл бұрын
convolutional and LSTM neural networks are the cutting edge for these applications
@Golde2Good
@Golde2Good 7 жыл бұрын
You should explain core parking in the near future.
@angelstrife
@angelstrife 7 жыл бұрын
Hi! Could you do a FPS 1%low explaination? I have seen so many tech reviewers use this term but i have no idea what it means.
@sniperunrepeat752
@sniperunrepeat752 7 жыл бұрын
Long Nguyen Games tend to have "stutters" (i.e. briefly running out of VRAM on say, a 1060 3gb) which can temporarily bring the minimum fps incredibly low. So 1% lows are used. All they mean is the minimum fps that doesn't factor in the bottom 1% of frames, to give a more realistic minimum
@Bayonet1809
@Bayonet1809 7 жыл бұрын
Could also be called the 99th percentile.
@unguidedone
@unguidedone 5 жыл бұрын
we need a firefox plugin that will log what youtube upload has paid promotions, skip past it and end the video when teh promotion happens. this video is an example of native advertisting
@mickeyhage
@mickeyhage 7 жыл бұрын
OCRs font work ive tried them but they dont properly. They dont read encrypted documents they spit out random incorrect letters.
@donaldfilbert4832
@donaldfilbert4832 7 жыл бұрын
OneNote has a pretty good built in OCR for small text articles - and it's free !! ABBYY FineReader does an excellent job converting image PDFs into searchable text based PDFs !!
@NineToFiveGamer
@NineToFiveGamer 7 жыл бұрын
I used to use an augmented translator app for my French tests. Shit just about worked half the time
@Quack201
@Quack201 7 жыл бұрын
So I guess the real question here is why is Luke only wearing socks while recording this? Doesn't Linus give sandals to all the employees?
@johneygd
@johneygd 7 жыл бұрын
But can OCR ever distinguich hand written numbers and letters from eachother? Such as 0's & o's, G's & 6's, 1's & i's ,H's & 4's , j's & i's, 7's & 1's ,0's & 8's etc,,,, because numbers and letters looks similar to eachother.
@araddadi2
@araddadi2 5 жыл бұрын
Watching this 10 minutes before class because I have a home and I’m a highly functional college student
@DanRobards
@DanRobards 7 жыл бұрын
Man, the ACR was great. Hardly any recoil
@DeppImAll
@DeppImAll 7 жыл бұрын
I mean tbh ... when I write in OneNote some text and microsoft can figure out what I just wrote and convert it into real characters I'm always astonished since my handwriting is horrible.
@bassmickey
@bassmickey 7 жыл бұрын
Funny used OCR last night. What a coincidence
@thornejman6467
@thornejman6467 7 жыл бұрын
Thumbs up if anyone else checked the videoquality at 0:36 xD
@rinoy_43
@rinoy_43 7 жыл бұрын
I've tried Tesseract. Its free and pretty accurate.
@todddembsky8321
@todddembsky8321 7 жыл бұрын
Luke, you have to tell me when you go on tour -- I need to leave the country at that point....
@GroovingPict
@GroovingPict 7 жыл бұрын
do you like dags?
@Juiceman777
@Juiceman777 2 жыл бұрын
I couldn't help but to think of the line from the movie Snatch when Brad Pitt said "ya like dags?" lol
@Mihnea729
@Mihnea729 7 жыл бұрын
Interesting !
@joerider5063
@joerider5063 7 жыл бұрын
Do speech recognition as fast as possible please.
@stayprofessional2453
@stayprofessional2453 7 жыл бұрын
Make an episode on network topologies
@terrybell898
@terrybell898 7 жыл бұрын
Micky: Ya like dags? Tommy: Dags? Micky: Yea, dags Tommy: OH, dogs, sure I like dags
@182ndNegociator
@182ndNegociator 7 жыл бұрын
What if it's supposed to say dag, that's also a completely legitimate word used in Australian English, plus it could also be used to describe a Directed Acyclic Graph, also known as a tree.
@MrEsChannelYT
@MrEsChannelYT 7 жыл бұрын
d'ya like dags?
@Exploreyourlife88
@Exploreyourlife88 3 жыл бұрын
Thanks
@metashrew
@metashrew 7 жыл бұрын
If the software were dutch, the word would be "dag" (which means day in english), and not "dog".
@feni_1553
@feni_1553 2 жыл бұрын
Images in video editing?
@levingthedream
@levingthedream 7 жыл бұрын
Is there any awesome free software that do this? Linux or PC. Besides Google drive that is
@pikotechsolutions
@pikotechsolutions 2 жыл бұрын
awesome
@_Disi
@_Disi 7 жыл бұрын
What about if you're trying to copy the line "D'ya like dags?" from Snatch?
@UNPhantom93
@UNPhantom93 7 жыл бұрын
Would be much better if was a fold able or detachable at least to use it as a tablet
@ThePiGuy24
@ThePiGuy24 7 жыл бұрын
I WANT INTERPRETIVE DANCE TRANSLATOR NOW!!!
@MrTuffarts
@MrTuffarts 7 жыл бұрын
Dag is a word OCR software would not pick this up spellcheck does not pickup this also
@blingerang
@blingerang 6 жыл бұрын
3:33 dag is actualy morning in dutch
@AndyPhu
@AndyPhu 7 жыл бұрын
This isn't in 4k! :(
@sahotaquack1
@sahotaquack1 7 жыл бұрын
Oxford Cambridge RSA
@Seag-Gaming
@Seag-Gaming 7 жыл бұрын
Who else had nostalgia @ 0:36?
@SuperManitu1
@SuperManitu1 7 жыл бұрын
Tesseract is the best OCR program out there. It is Open Source and runs on all major OS
@94213915
@94213915 5 жыл бұрын
How can I run it on Windows
@1OldWriter
@1OldWriter 7 жыл бұрын
Techquickie you do know most scanning software do this as part of their operation. If your's doesn't perhaps you should get a new one.
@nitini.764
@nitini.764 6 жыл бұрын
I liked this "don't worry, be happy" in your video. Are you a Meher Baba lover too!!!!
@marcusleung8985
@marcusleung8985 7 жыл бұрын
what about Fourier transform?
@bas116677
@bas116677 7 жыл бұрын
Dag actually means Hey or day in Dutch!
@kdm_6799
@kdm_6799 7 жыл бұрын
Bas Roelofs dag means bye too
@7EEVEE
@7EEVEE 7 жыл бұрын
most scanners look pretty good to be fair
@isabellaereshki
@isabellaereshki 7 жыл бұрын
I liked your dancing, ignore dennis. great video.
@BenPotts
@BenPotts 7 жыл бұрын
Nice dancing, Luke
@Rembo2662
@Rembo2662 7 жыл бұрын
tfw dag is day in dutch
@megabithero
@megabithero 7 жыл бұрын
My Galaxy S3 could do this. Made lab reports super manageable.
@svsrkpraveen
@svsrkpraveen 6 жыл бұрын
When did Dan Reynolds start doing tech stuff?
@Aaron-jv7pc
@Aaron-jv7pc 7 жыл бұрын
Luke kinda reminds me of Chris Pratt
@Brusanan
@Brusanan 7 жыл бұрын
Not one mention of neural networks?
@zcuipylo
@zcuipylo 7 жыл бұрын
TPS reports!!!!!! What a perfect example. Almost an easter egg.
@Shirojm
@Shirojm 7 жыл бұрын
So use a normal "photographic" scanner , then use OCR services such as google drive .