A-Z | 0-9 Handwritten Character Recognition Crop Tool Using Python | Complete Tutorial (English)
Vložit
- čas přidán 16. 11. 2021
- In this tutorial you will learn to implement a real time 'A-Z and 0-9 Handwritten Character Recognition Crop Tool' using Python and related modules such as Tensorflow, Keras, OpenCV and Numpy. This tutorial
has covered with 3 main sections such as Data Pre-processing, CNN Architecture and Real Time Application Implementation.
In data pre-processing we are going to merge two datasets, which consist of both A-Z handwritten characters and 0-9 handwritten digits respectively. Then the individual merged dataset is going to be fed in to our CNN and we are going to pick the best model which will be created at the end of our model training phase. Then we are going to embed our trained CNN model to real time character cropping application and go through with interesting results.
Chapters/Sections
=================
00:01:52 - Data Pre-processing
00:27:00 - CNN Architecture
00:40:00 - Real Time Crop Tool Application
Download the project
==================
Github:
github.com/Sinsank/A-Z-0-9-Ha...
Google Drive: drive.google.com/drive/folder...
References
==========
Kaggle A-Z handwritten dataset: www.kaggle.com/sachinpatel21/...
MNIST dataset: yann.lecun.com/exdb/mnist/
Social Media
===========
Facebook: / sintax.tech.blog
Linkedin: / sineth-sankalpa-9aa4331ab
Subscribe 'The Sineth' and hit on the bell icon.
/ @thesineth
Thanks for watching ❤
00:01:52 - Data Pre-processing
00:27:00 - CNN Architecture
00:40:00 - Real Time Crop Tool Application
speechless සිනේ!
ශුභ අනාගතයක්! ❤️❤️
Proud of you bro ♥️♥️
ජය වේවා 🔥
niyamai sir.....
Thanks 😊
ආඩම්බරයි සහෝුදරයා...
Nice ❤️
💪💪ජය වේවා
Super bro❤️
tq bro❤
❤️🔥
❤️
brh❤️
👍🤗
While training the model it is giving error like ValueError: Data cardinality is ambiguous: x sizes: 88490
y sizes: 353960
Make sure all arrays contain the same number of samples
😎😎🍃
how to dowload the training and testing labels?
can you please tell the learning rate of this model
Sir, can you please explain the use of line 10 of code i.e digit_target+=26 ,,, I'm unable to figure out use of that line
Basically we are adding 26 to each and every label of digit dataset. Since we are appending a handwritten digit dataset to the handwritten character dataset, the final label of character dataset should be 25. (From A to Z there are 26 letters) From Z to onwards there should be digit labels. Digit label dataset should be started with label 26. But the original digit dataset labels are initialized from 0. That should be initialized from 26. That's why it is adding 26 to each label present in the digit dataset.
Hope you have been cleared with the doubt.
In the application code I am getting this error in last
NameError: name 'setup_display' is not defined
Please help
Can you tell me what are the packages that needs to be download to run this project?
(Roughly)
Tensorflow
Keras
Numpy
Pandas
OpenCV
Scikit learn
hello, i want to ask you if you can help me when writing a character directly recognize it and not crop it
You have two options. First one is you can train your OCR model by passing images with their bounding rectangle co-ordinates. In here you have to go through all the images in the dataset and find out the x,y,w,h co-ordinate points of each and every image present. And finally you can train out your CNN pipeline according to the need.
The second option is going through the approach of transfer learning. Using transfer learning means train a pre-trained CNN according to our need.
Bro I am getting error at the last part of application
-------> cv2.setMouseCallback(window_name, mouse_click_event)
error : OpenCV(4.1.2) /io/opencv/modules/highgui/src/window_QT.cpp:714: error: (-27:Null pointer) NULL window handler in function 'cvSetMouseCallback'
above is the error I get
please reply me what can I do now
I was trying in google colab
Plz help me as fast as possible
Did you google it and have a try?
Can use this code for word recognition ?
no you have to add an RNN
This is only for character recognition.
Bro I am getting the error of unable to allot 1.3 GB of space. Kindly look into this
This is because of lack of space of your current RAM to load your dataset perfectly. Try to upload your dataset into a google drive and train your CNN using Google Colab if you have no any chance to continue your training progress with another computer.
bro where is the training dataset(numpy train dataset and labels)
You have to downoad the dataset and locate in your project directory or any local location of your computer. Then copy out the location url and paste it inside implementation.
Do you got it? Can I know how to download the test and train labels
I am not able to load the A-Z CSV file in my notebook. What to do
Hi,
If you have already downloaded the CSV, all you need to do is correctly specify the respective file path and get executed.
Now it shows data mismatch in some some columns even though I type .astype('float')
@@kavipriyaravikumar5136 Checkout this: www.rasgoml.com/feature-engineering-tutorials/how-to-handle-data-type-mismatch-in-pandas
can I contact you to help me build the project on my desktop ?
hello ?
Sure. Check out the description to contact me.
Bro how i will contact you to help me build the project.
Send me a dm via whatsapp (+94 77 3283048)
can i use my own dataset?
Yes you can. But the CNN architecture should be customized according to the distribution of your dataset.
how do i run through ur git code directly
First download the repository of project materials from the github (link is in the description). Then you can execute your downloaded application successfully. Keep in mind to download and install all the required module packages/libraries before the execution.
@@TheSineth thanks mann ❤️❤️
Application ko close karne ke liye ham kis button ka use karenge
@@nikitasingh8818 Can you please send this message again in English ?
Escape key
@@TheSineth Do you have a report of this project, we have to do it for college.
@@TheSineth okky 😊
@@TheSineth thank you so much
Can you send the soft copy
Checkout the description to download all the project materials. BTW you can convert jupyter notebooks into pdf format.
👩💻🥺
i got this problem :
plt.imshow(shuffled_data[i].reshape(28, 28), interpolation="nearest", cmap="gray")
9 plt.show()
TypeError: 'NoneType' object is not subscriptable
❤️