Image Classification

Question 1

What can the convolutional process be considered as?

Answer 1

could be considered as acting an image convolution and the convolutional layer is the convolution output of the previous layer

Question 2

Where does the trainable filter lie?

Answer 2

from the input layer to hidden layer

Question 3

What could the trainable filter be considered as?

Answer 3

a feature map with shared weights and bias

Question 4

How is a DCN-like architecture trained?

Answer 4

trained in an unsupervised manner to learn hierarchical sparse features that were locally invariant to small shifts and distortions

Question 5

What do DCN’s use?

Answer 5

3 or more convolutional layers where each layer is responsible for extracting one or more feature of the image.

Question 6

What is max pooling?

Answer 6

Max pooling is used between convolutional layers to reduce the dimensions of an image by taking the maximum pixel value of a grid

Question 7

Why is max pooling useful?

Answer 7

reduce overfitting and makes the model more generic

Question 8

What other layers to DCN’s have after convolutional layers?

Answer 8

connected layers between input and output layers to ensure the dimensional outputs are correct

Question 9

What do convolutional neural networks consist of?

Answer 9

Convolutional Layer, Pooling Layer and Fully-Connected Layer

Question 10

What are the 2 main process of CNN’s and where do they occur?

Answer 10

convolution and sampling, which will happen on convolutional layers and max pooling layers

Question 11

What is convolution?

Answer 11

Every neuron takes inputs from a rectangular n × n section of the previous layer, the rectangular section is called local receptive field

Question 12

Why can the parameters of CNN training equation be viewed as a trainable filter?

Answer 12

Since the every local receptive field takes same weights wr,c and biases b from the equation above, the parameters could be viewed as a trainable filter or kernel F

Question 13

Where does the sampling process happen?

Answer 13

between convolutional layer and pooling layer

Question 14

What does the pooling layer do?

Answer 14

takes small rectangular blocks from the convolutional layer and subsamples it

Question 15

What can we consider as the output to the pooling layer?

Answer 15

the maximum of the block as the single output to pooling layer

Question 16

What can follow a convolutional layer?

Answer 16

a pooling layer

Question 17

How is a CNN used to classify images (input output)?

Answer 17

The input layer of the network contains neurons encoding the values of the input pixels. The output layer of the network contains a number of neurons which stand for each type of image label.

Question 18

Why are CNN’s useful for image classification?

Answer 18

CNNs have much fewer connections and parameters, and also, they are easier to train. Discarding the fully connected strategy means to pay more attention on the regional structure, which is very meaningful when we take image processing into consideration, since there are less relations between different region of an image