Skip to content

EgOrlukha/Image-classification-using-Neural-network

Repository files navigation

Image-classification-using-Neural-network

Its A Basic Image Classifier using Keras we can the structure i use is similar which is available in keras documentation

Layer (type) Output Shape Param #

conv2d_33 (Conv2D) (None, 148, 148, 32) 896


activation_31 (Activation) (None, 148, 148, 32) 0


max_pooling2d_29 (MaxPooling (None, 49, 49, 32) 0


conv2d_34 (Conv2D) (None, 47, 47, 32) 9248


activation_32 (Activation) (None, 47, 47, 32) 0


max_pooling2d_30 (MaxPooling (None, 23, 23, 32) 0


conv2d_35 (Conv2D) (None, 21, 21, 64) 18496


activation_33 (Activation) (None, 21, 21, 64) 0


max_pooling2d_31 (MaxPooling (None, 10, 10, 64) 0


flatten_3 (Flatten) (None, 6400) 0


dense_2 (Dense) (None, 64) 409664


activation_34 (Activation) (None, 64) 0


dropout_1 (Dropout) (None, 64) 0


dense_3 (Dense) (None, 1) 65


activation_35 (Activation) (None, 1) 0

Total params: 438,369 Trainable params: 438,369 Non-trainable params: 0

explaination?-:

  Image ->  150*150*3 { height -> 150, width->150, depth or RGB -> 3}
  than i apply filter 32 3*3 =>  (148,148,32)
  where 32 is stack of filter basically we found 148 unique position while doing convolution
  =>(I-F)/(S+1)
  =>total number of observation = (3*3*3)*32 -> 864

what is convolution?

  it is basically moving the filter  over the image with some stride 
  we do dot product of w and x value which lie in the filter and sum it up which give us a single
  value this process is called convolution
  its basically (w.transpose().x +bais)

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published