/Malayalam-Character-Recognition

Malayalam Character Recognition using CNNs in python

Primary LanguageJupyter Notebook

Malayalam-Character-Recognition

Refer to Project-Notebooks folder for updates

Project-Notebooks
=================

  • Dataset_Preparation : Converts images to 32X32 Grayscale images with adaptive thresholding
  • NP-Dataset : X.npy and y.npy are numpy arrays of (34553,32,32) & (34553,) dimensions
  • Training_Dataset_Generation: Functions to Reshape and Normalize data.OneHotEncode Labels etc
  • Models/ : Folder which contains weights of trained Models
  • CNN_Model : Model is designed here. Entire training pipeline happens here. Weights are saved in the end
  • Prediction_Estimation : Generated the Prediction. Used for performance evaluation,Plotting etc



    • Training Results
    ================

    Accuracy & Sample Predictions
    ============================
    Trained the model using floydhub service
    Results are below :





    Model Used
    ==========



    Deployment
    ==========



    Coming Soon
    ===========

    1. Adding more data + augmentation
    2. Deployment in a web-app