Hindi Handwritten Character recognition using Constitutional Deep neural networks

Implementation Details:
1. Execute load_image_into_pickle.py file
a. It takes input of images from folder /Dataset
b. Image pre-process and feature extraction done
c. Saves output file in folder “dataset_pickles” as dataset_46_classes.pickle
2. Execute train_model.py
a. Gets input file dataset_46_classes.pickle
b. Convolutional deep neural network used for training model
c. output of train model is stored in folder /out
3. Execute predict_model.py
a. It takes input of test images from folder /test_images
b. Model compares with the weight model created during training “hhrc_nn.h5py”
c. Predicts output

Python demo

SMS spam detection using Recurrent Neural Networks

Implementation details:
1. SMS dataset is taken, which is represented with ham and Spam
2. Dataset is pre-processed
a.removed all stop word, stop words are given in separate file
b. converted type of sms to numeric ham-1 and spam-2
3. Pre-processed dataset is tokenized, applied pad_sequnce, then converted to vectors
4. Then they are trained by network model and creates output file h5, json
5. Test input is given to the model and predicts the output
6. Comparison and accuracy are arrived.

Python Demo