Image_classifier_using_Keras

Image Classification using Convolutional Neural Nets and Keras. This classifier uses LeNet-5, a pioneering 7-level convolutional network by LeCun et al in 1998 that classifies digits.

Requirements: Python 3.5 ,Keras 2.0.2 , Tensorflow 1.2.1 , OpenCV 3.3, numpy 1.11.0

How to:

Train your own model:

 $ python train_network.py --dataset images --model Bharath_Kumar.model   

Test your own model:

 python test_network.py --model Bharath_Kumar.model --image images/examples/bk19.jpg    

Add Your own Dataset:

 Add your training images in  images/<Your Label>/

Contents /Scripts of Barath Classifier :):

-train_network.py :

    To train the  model.    

-test_network.py:

    To train the  model.    

-Bharath_Kumar.model :

    Model generated during thhe training.    

- data.zip :

    Contains Images for Training. Use your owm data set here.    

-BK_lenet.zip :

    LeNet Architecture.      

About The Model:

Trained using Backpropogation algorithm with stochastic gradint descent. This is a Binary Classifier.

Accuracies after 25 epochs:

For classification based model:

-Train acc: 96.4665%    
-Test acc : 88.5039%