Text detection with Python and Opencv | OCR using EasyOCR | Computer vision tutorial

[@gloryness]

I noticed that test2.png wasn't a perfect detection because "$5.00 FEE" was picked up as "S5.00 FEE" Aside from that, really helpful tutorial!

[@mermoqqueijo]

Hey, just found out about your channel and is very good! I am also a Computer Vision practitioner, and I will use your code for an application with a robot focused for the pedagogy. Basically, the goal is to detect Blue, Red and Green texts and it send an information to the robot to pick it up a piece of that color =). I also saw that you have a tutorial for the color detection, so I might use that also haha. Anyways, great job!

[@ComputerVisionEngineer]

Sounds like a very cool project! 😃🤖 Thank you for your support! 🙌

[@WelcomeToMyLife888]

Thanks! Your videos helped me a lot in trying to get into computer vision.

[@ComputerVisionEngineer]

☺️ my pleasure! I am happy to help you in your computer vision journey! 😃💪

[@thatglamgirl31]

Thank u sooo much it helped a lot

[@gabrielchinzz3332]

extremely helpful tutorial for a project, im doing right now related to OCR.Only problem now is that this model tends mistake the $ sign for an S when testing some images,other than that it is able to detect texts quite well!

[@sjr787]

Hey can I know what project are you doing

[@gabrielchinzz3332]

@@sjr787 Ocr scanner to scan a food pricelist and update the database

[@pratyusha18]

Hey great and to the point explanation. One question: when implementing this I am getting this error - UnicodeEncodeError: 'charmap' codec can't encode characters in position .... Any suggestions on how to handle this?

[@tribagussoleman]

thank you I hope god bless you because your video very usefull

[@Muhamed_Esm3iL]

very goood our ENG

[@RWM85]

Thanks, great explanation.

[@eldea9745]

hey i have question. can easyocr recognize text with white background only?

[@MrMilesfinn]

excellent. Great organization.

[@ComputerVisionEngineer]

Thank you! Glad you enjoyed it! 😃

[@hamzawi2752]

Thank you very much. It is informative.

[@ComputerVisionEngineer]

I am happy you find it useful! 😊🙌

[@greyswandir2807]

I'm trying to write a code for selecting an area of the screen to scan for text (using tesseract) but it's going horribly so far. Please help :(

[@nikowolfwood]

Hi!! I found your channel today, I'm trying to make a similar functionality but with text in real time that appears on screen (to make translations in real time). Do you think that is possible using opencv or I will need to use another image detection model? Cheers! and keep up with the good work!

[@ComputerVisionEngineer]

Thank you! I think you could do it with EasyOCR, the same way I do it in the video. Let me know how it goes! 😃🙌

[@nikowolfwood]

@@ComputerVisionEngineer Hi I just wanted to tell you that I managed to make the program with easyOCR + googletrans + tkinter to display text on screen, but the problem is that the process of OCR+translation takes way too long (like 15 seconds to translate 20 words) so Ill keep researching for a faster way.

[@ComputerVisionEngineer]

​@@nikowolfwood Cool! Thank you for sharing the update on your project! 💪

[@oliviaonyinye595]

@@nikowolfwood can you please share your codes . I am working on something similar but I am having issues

[@manifestasisanubari]

@@nikowolfwood It would be nice to see a demo video of you. It sounds like Google Lens (Translate). Kudos!

[@philipgeorgiev]

Dude , you rock!

[@ComputerVisionEngineer]

Thank you for your support! 🙌

[@thaihoanguyen]

my output using print(Text) is only "Using CPU. Note: This module is much faster with a GPU". Nothing else. why? Help me please

[@SuzanBawani]

i also get the same output

[@NicoleAgyapong]

Me too

[@brahimaissa]

Thanks for you for this, i hope if you can to explain how it works.

[@stern7658]

hey buddy could you do an update on the code that have it print out the text to the consol and also lets say that some words letter has a font that is bgger and bolder than the rest that it also target that too then print out tit to the console, example text that is bigger or like a header it print it like this for example: since caution is more bigger it it print H1- CAUTION for caution, H2- maintain social destancing for maintain social destancing etc. you get the point

[@hemasripidintla-jg4ss]

Can you do a video on vertical text detection For eg: 1 6 7 E 5 Text should detect and recognise vertically also Pls make a video on this

[@ComputerVisionEngineer]

Hey Hemasri, that is a very good idea! I will look into it for a future video. 🙌

[@danilzubarev2952]

Thank you :)

[@ComputerVisionEngineer]

You are welcome! 😃🙌

[@muzhafarhashim1234]

Great tutorial ! Can we add language translation as an output for this project ? Is like image to text translation .

[@ComputerVisionEngineer]

Thank you! Sure, that is perfectly possible! It is just a matter of plugging the output with a text translation API. 😃🙌

[@muzhafarhashim1234]

@@ComputerVisionEngineer can you do a video about that ? im a python beginner or you can send me a coding with the translator

[@ComputerVisionEngineer]

Sure, I liked the idea for a tutorial, I can make a video about it. 😃 I am working in other projects right now, but I will find some time to do it within the next few weeks! 💪💪

[@PhatTran-cp8yf]

can I export to excel after concerted in python

[@shreyasskrishnasa637]

Can we convert this model running from GPU to CPU and find its refresh rate and how?

[@kwelsty]

wait what software does this guyu even use? i dont even know how to start, could you also make a video explaining how to download everything from scratch?

[@vasilisvasileiou8406]

hello sir!, Is it possible to train an easyOCR model to detect images based on my own dataset?

[@philipgeorgiev]

Is there anything better than easyocr that we can run locally like claude 3 haiku ?

[@sayisha2472]

Nice tutorial! Is it possible to read text from an inverted image using easyocr and that the output is expected to come as expected?

[@ComputerVisionEngineer]

Hey, it is much better if the text is aligned. But take a look at easyocr documentation, the parameter rotation_info from the function readtext may be what you are looking for. 🙌

[@middlelevelgamer9180]

Is there a way to group high score text and low score text and display only high score text?

[@soroushlotfimehr]

TNX man 👨 ❤❤❤❤❤❤

[@harishdasari2641]

Hi I'm getting error like " process finished with exit code -1073742795 (0xC000001D) while executing easyocr with python 3.9. can provide any solution for this issue.

[@SukumarReddyMandem-et6gd]

Hi Your videos are helpful...Can we build OCR model to detect date and time in the video. For me it will be in the same place throught the video. Please help me.

[@ComputerVisionEngineer]

Hey, yes sure it is possible. You can use opencv to read frames from the video, then crop the frames where the date / time is located and input the crop into easyocr. 💪

[@SukumarReddyMandem-et6gd]

@@ComputerVisionEngineer Thank you so much for your reply. I implemented same but it's detecting differently each time. For example if date and time is 05/05/2023 07:08:23 some times it is detecting correctly remaining all times it is detecting as 07.08.23, 077778.23, 27.77.28 something like this each time. Can we fine tune to detect correctly. Format will be exactly same for each video as camera used to record footage is same. Please help me.

[@satypk8664]

In my case it is detecting & writing only 3rd line "AND BIKE USE".

[@ComputerVisionEngineer]

Are you installing the same easyocr version? Are you using the same images I used in the video?

[@satypk8664]

@@ComputerVisionEngineeryes , followed the same code also , for drawing rectangle i modified by using loop so finally was able to draw rectangle around everything .but in case of writing issue is still there

[@holdengreene9717]

Ive had a lot of trouble just getting easyOCR installed.I keep getting errors when the depoendencies are getting installed, specifically pytorch. Ive read that pytorch is not compatable with any python version after 3.7 which i find odd and also this is never mentioned in the video. This feels like it should be a simple pip install comand but i have been trying for hours and nothing works. Anyone else encoutnering this issue?

[@ComputerVisionEngineer]

Hey, I just cloned the repository and everything works properly on my end. I am using Python 3.8, on a Linux system. I added a requirements.txt file to the repo with all the versions of the different packages in case it helps. Let me know how it goes! 🙌

[@holdengreene9717]

@@ComputerVisionEngineer Nope. Still get the following C:\Users\hgree>pip install easyocr==1.6.1 Collecting easyocr==1.6.1 Using cached easyocr-1.6.1-py3-none-any.whl (2.9 MB) Requirement already satisfied: torch in c:\users\hgree\appdata\local\programs\python\python311\lib\site-packages (from easyocr==1.6.1) (2.1.0.dev20230306+cpu) ERROR: Could not find a version that satisfies the requirement torchvision>=0.5 (from easyocr) (from versions: 0.1.6, 0.1.7, 0.1.8, 0.1.9, 0.2.0, 0.2.1, 0.2.2, 0.2.2.post2, 0.2.2.post3) ERROR: No matching distribution found for torchvision>=0.5

[@holdengreene9717]

this is odd since torchvision is installed.... Requirement already satisfied: torchvision in c:\users\hgree\appdata\local\programs\python\python311\lib\site-packages (0.2.2)

[@holdengreene9717]

@@ComputerVisionEngineer It looks like the issue is that easyOCR wants a version of torchvision > 0.5 but when I manually install I cant seem to find a version above 0.2.2 PS C:\Users\hgree> pip3 install torchvision ERROR: Could not find a version that satisfies the requirement torchvision (from versions: 0.1.6, 0.1.7, 0.1.8, 0.1.9, 0.2.0, 0.2.1, 0.2.2, 0.2.2.post2, 0.2.2.post3) ERROR: No matching distribution found for torchvision

[@oguapi]

🙌

[@ComputerVisionEngineer]

😄🙌

[@RabeeQasim]

can you do the same tutorial with yolov8?

[@ComputerVisionEngineer]

I will try to. 🙌

[@RabeeQasim]

@@ComputerVisionEngineer thank you

[@hayder_khalid]

hi that good job but i need using Blurred image

[@letlhogonolosetlhodi7494]

The word "but" disqualifies what you said before that. Sometimes it's a great idea to thank somebody for doing a great job especially when they are not obligated to do it. Then in a separate conversation we can't ask them nicely if it's possible to use blurred images as a use case.

[@hayderalaasdy8611]

@@letlhogonolosetlhodi7494 I am sorry i not good in english

[@ComputerVisionEngineer]

Thank you so much Haider and Letlhogonolo for your contributions. Haider, do you mean you need to read the text from a blurred image?

[@hayderalaasdy8611]

@@ComputerVisionEngineer yes

[@krzysztofbiay462]

@@letlhogonolosetlhodi7494 the word "but" does not disqualify anything, LOL. It just introduces contrast or additional information to what was said before. I have no idea who educated you... but he sure as hell owes u a refund. And I have no idea what made you think that you are welcome to interject here and scold other people. Because no one cares about your opinion so you are well-advised to take your autism elsewhere. Also - yes, he was not obligated to do it but it's not like he did any charity. KZbinrs make videos to develop their channels and, hopefully, to get paid. And if they don't do well, users have like 10 000 other videos to watch so it's up to channel to deliver what consumer needs and then he can be thankful that anyone watches. Last but not least, he did a decent job but by no means "great" - many people, including me, had issues with easyocr's installation. Nowhere it was mentioned that nvidia-cuda-toolkit needs to be installed. And it needs to. And there were some other problems.