Bangla Handwritten Character Recognition Using CNN
Vložit
- čas přidán 14. 08. 2020
- I have developed a python library named imgclassifier on top of PyTorch which requires only one line of code for classification. Follow the attached notebook entitled imgclassifier_character_recognition for Banga Handwritten Character Recognition using imgclassifier.
*imgclassifier library: github.com/mehedihasanbijoy/i...
*imgclassifier_character_recognition: colab.research.google.com/dri...
*BHCR repository: github.com/mehedihasanbijoy/B...
Hi! Thank you for the video. This is very helpful!
Namoskar, sir apni input training labeled dataset ta kotha theke nilen? Kaggle?
Juktakhor er jonyo ki kore egobo, jodi short e bolen khub valo hoi
vaia nice tutorial tobe ektu explanation korle valo hoto
Hi, I'm getting "AttributeError: module 'keras.preprocessing.image' has no attribute 'load_img'" error in the draw_n_guess_the_character() part, how can I resolve it?
how can I post an image as input for checking against the model?
Have you uploaded the code in github?
Hi vaiya .your video is amazing. I was wondering that if we work with complex letter then is this code working? If you give mail or any contact information then it will be easier for me . Because I am interested in Real-Time Bangla Handwritten Characters and Digits Recognition using Adopted Convolutional Neural Network this topic.i have lots of questions 😅
It may requires few modifications in code to connect your dataset to the model. And in case of model, you may try adding some hidden layers, different sized filters, more no. of filters hence different number of filter channels, different stride number, different activation functions, and so on. However, you can ask me remaining questions at mhb6434@gmail.com
what is the dataset?? Is the dataset is handwritten or printed??
What if i want to import my own dataset how will i make it
which dataset is suitable for to predicting english handwritten text using this model.
the MNIST Dataset
I was trying to find a software that translates doctors prescription...can you help me develop it?
Could you please email me the details at mhb6434@gmail.com?
Do i need GPU for training this model? And is google colab enough for training 50 classes dataset? It will be good to see your reply. thank you
Yes, colab is enough for training a model on large-scale datasets, but you need to be patient. For instance, it may require hours to complete training for each epoch, so I would suggest saving the model after each epoch.
@@mehedihasanbijoy6609 ❤️
Where will i get the dataset sir? What is the name of the dataset?
drive.google.com/file/d/1dcs1a7Yt9onWlbxrNBOkT7HNC6VWapS2/view?usp=share_link
Can you tell me where from the data is being taken? I mean what is the source link of your dataset?
the name of the dataset is CMATERdb. I collected it from the internet.
@@mehedihasanbijoy6609 Thanks a lot 😊
What is the name of the dataset? Is it CMATERdb?
yes it is CMATERdb
Bro how to this for Kannada language. I got data set in png file
0. clone the repository and have a look on the format of the dataset, and convert your dataset into this format.
1. make few modifications in the code(particularly in dataloader) according to your dataset's format.
you can do it in either way.
Thank you bro.
What is the solution ?
classifier.fit_generator(training_set, steps_per_epoch = 12000, epochs = 10,
Epoch 1/10
375/12000 [..............................] - ETA: 25:23 - loss: 2.7628 - accuracy: 0.2730WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 120000 batches). You may need to use the repeat() function when building your dataset.
WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 3000 batches). You may need to use the repeat() function when building your dataset.
I just removed the steps_per_epoch part and let it run
@@chiefsenpai969 it worked?
@@alarafat6859 Yes
Have you uploaded the code in github?
yes, you will find the github repository link in the video description.
@@mehedihasanbijoy6609 How can I gate other files?image?