Convolutional Neural Network (CNN) Steps :
- Input Image
- Convolution Layer (<– apply Activation Function ‘relu’)
- Max Pooling
- Input for ANN
- Fully connected Hidden Layer(s)
- Output (output_dim=1)
- For data augmentation use image preprocessing
For 2 Classifications use binary cross entropy
For more than 2 Classifications use – categorical cross entropy
- To improve accuracy of model either you can add Convolution Layer
- You can also try by increasing target_size (ex target_size =(64, 64 ) -> target_size =(128, 128) of both training and test Set.
Gradient descent : For finding the minimum of a function Fully connected Hidden Layer.
- Stochastic gradient descent
Example Model –