Image Classification using Convolutional Neural Networks - saidu941/Deep-Neural-Networks GitHub Wiki

About the Project:

Classifying the image by implementing the Convolutional Neural Networks concept, providing the dataset with the images of A and images of B, eg. say dogs and cats.

With 80:20 split of data, We input the training data with 8000 images of both the dogs and cats with including the labels and after training the model, we then test the output by providing the test data with 2000 images of both dogs and cats as input to the predicted CNN model to get the desired classified output.

Workflow:

• Importing the Keras library (Keras is a high-level neural networks API, written in Python and capable of running on top of TensorFlow, CNTK, or Theano.) More advanced than the TensorFlow library

• Step 1. Convolution: on the set of images with the Feature Map

• Step 2. Max pooling: (pooling by term means reducing the size without losing the performance of the feature maps from the convolution step and saving them as Pooled Feature maps). Max pooling takes the maximum of the number in the feature maps by traversing along the feature maps. We call this as pooling layer. we are reducing the complexity by preserving the performance

• Step 3. Flattening: