AYUSH-ISHAN/Devnagri-Letter-and-Digit-Recognition

This is the code of Handwritten Hindi Recognition system using Deep Learning and more precisely using Computer Vision. # Devnagri-Handwritten-Recognition This is the code of Handwritten Hindi Recognition system using Deep Learning and more precisely using Computer Vision. # SOME IMPORTANT POINT: It can be noticed that when you open the main file (MOSAIC_try.ipnyb file) it seems there is an error in the code but don't worry about it the code is fine and it works .. # THE DATASET: I used the dataset provided by Kaggle whose link is: Dataset Some reference image are added here.. And also some of the other image which it predicted correctly. But feel free to use your dataset... # HOW TO RUN ON YOUR TRAINING AND TEST SET: Well, in the code it has provided the option to give the location of folder of your test and training set. By provivng the correct directory it will work.. But the important point is if you have to make it predict on your customise letters, then there are two ways to do so.. 1. Include the format of letters on which you are going to predict in the test and train dataset (CLEVER WAY !!) 2. Now, the second is to do some image preprocessing by the use of OTSU Thresholding, noise removal and other similar process. # THE IMAGE AUGMENTATION PART: The image augmentataion of the image worked fine for me under the mentioned arguments values but it may be different for you..So, feel free to do some experiment. # THE IMAGE PREPROCESSING PART: I am adding here the reference for the image preprocessing part. The main notion of this project to create an univerasal recognition system which works for all types of Devnagri Letter. # HERE ARE SOME PLOTS TO SHOW THE ACCURACY ACHIEVED : # NOW, THE FINAL DATABASE TO SHOW THE EXACT ACCURACY AND LOSS: The final Accuracy was 98.18 % on Training Set and 98.42 % on Validation Set.