Update README.md

Convolutional Neural Network/README.md

@@ -1 +1,77 @@
 # Artificial Neural Netwrok Vs. Convolutional Nural Network !! [Image Classification]
+
+## Notebook :
+
+  [Notebook](https://colab.research.google.com/drive/1cIIXFeROtJf7MovU7j6ipBDvur-X1o9M?usp=sharing)
+
+## Dataset :
+
+  [**CIFAR-10**](https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz). (In this project the dataset has been directly loaded from the Tensorflow datasets :
+    ```
+    tensorflow.keras.datasets.cifar10.load_data()
+    ```
+  The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes, with 6000 images per class. There are 50000 training images and 10000 test images.
+
+## Models :
+
+  * ***Net-A*** - A single layer perceptron with 10 outputs without any hidden layer
+     ```
+     _________________________________________________________________
+      Layer (type)                 Output Shape              Param #   
+      =================================================================
+      flatten_13 (Flatten)         (None, 3072)              0         
+      _________________________________________________________________
+      dense_23 (Dense)             (None, 10)                30730     
+      =================================================================
+        Total params: 30,730
+        Trainable params: 30,730
+        Non-trainable params: 0
+      ```  
+  * ***Net-B*** - Neural network with slightly more complication , that is , it has a hidden layer with 300 nodes and adds a non-linearity between the layers.
+    ```
+    _________________________________________________________________
+    Layer (type)                 Output Shape              Param #   
+    =================================================================
+    flatten_14 (Flatten)         (None, 3072)              0         
+    _________________________________________________________________
+    dense_24 (Dense)             (None, 300)               921900    
+    _________________________________________________________________
+    dense_25 (Dense)             (None, 10)                3010      
+    =================================================================
+      Total params: 924,910
+      Trainable params: 924,910
+      Non-trainable params: 0
+    ```
+   
+  * ***Net-C*** - A Convolutional Neural Network with one convolution layer , with 25 filtres of size 5x5 , with padding and activation function set to "relu". It is followed by                   a Max-pooling layer of size (2,2) with strides=2.Then after flattenning , a dense layer of 25 nodes is added before the final output layer with "softmax"                         activation.   
+    ```
+    _________________________________________________________________
+    Layer (type)                 Output Shape              Param #   
+    =================================================================
+    conv2d_9 (Conv2D)            (None, 32, 32, 25)        1900      
+    _________________________________________________________________
+    max_pooling2d_9 (MaxPooling2 (None, 16, 16, 25)        0         
+    _________________________________________________________________
+    flatten_15 (Flatten)         (None, 6400)              0         
+    _________________________________________________________________
+    dense_26 (Dense)             (None, 25)                160025    
+    _________________________________________________________________
+    dense_27 (Dense)             (None, 10)                260       
+    =================================================================
+      Total params: 162,185
+      Trainable params: 162,185
+      Non-trainable params: 0
+    ```
+## Visualisation :
+
+   A visualization of all the results with respect to the images and their classes are done , which gives an idea of the performance of each of the networks.
+   
+## Results :
+
+   Each of the networks are iterated through 50 epochs.
+   Test accuracy of the networks are given below:
+```
+    Test accuracy of NetA-  10.180000215768814 %
+    Test accuracy of NetB-  46.639999747276306 %
+    Test accuracy of NetC-  63.98000121116638 %
+```