Uses TensorFlow as backend for CNN and includes
- layers- parameter calculations- load keras and EBImage packages- read images- explore images and image data- resize and reshape images- one hot encoding- sequential model- compile model- fit model- evaluate model- prediction- confusion matrix
Embedded Files
C-Convolutional
N-Neural
N-Network
What is CNN ?
What is CNN ?
A Convolutional Neural Network (ConvNet/CNN) is a Deep Learning algorithm which can take in an input image, assign importance (learnable weights and biases) to various aspects/objects in the image and be able to differentiate one from the other. The pre-processing required in a ConvNet is much lower as compared to other classification algorithms. While in primitive methods filters are hand-engineered, with enough training, ConvNets have the ability to learn these filters/characteristics.
The architecture of a ConvNet is analogous to that of the connectivity pattern of Neurons in the Human Brain and was inspired by the organization of the Visual Cortex. Individual neurons respond to stimuli only in a restricted region of the visual field known as the Receptive Field. A collection of such fields overlap to cover the entire visual area.
Why is CNN used?
Why is CNN used?
In machine learning, Convolutional Neural Networks (CNN or ConvNet) are complex feed forward neural networks. CNNs are used for image classification and recognition because of its high accuracy. It was proposed by computer scientist Yann LeCun in the late 90s, when he was inspired from the human visual perception of recognizing things. The CNN follows a hierarchical model which works on building a network, like a funnel, and finally gives out a fully-connected layer where all the neurons are connected to each other and the output is processed.
Why is CNN better ?
Why is CNN better ?
Over the years, research on convolutional neural networks (CNNs) has progressed rapidly, however the real-world deployment of these models is often limited by computing resources and memory constraints. What has also led to extensive research in ConvNets is the accuracy on difficult classification tasks that require understanding abstract concepts in images.
Another reason why CNN are hugely popular is because of their architecture — the best thing is there is no need of feature extraction. The system learns to do feature extraction and the core concept of CNN is, it uses convolution of image and filters to generate invariant features which are passed on to the next layer. The features in next layer are convoluted with different filters to generate more invariant and abstract features and the process continues till one gets final feature / output (let say face of X) which is invariant to occlusions.
Also, another key feature is that deep convolutional networks are flexible and work well on image data. As one researcher points out, convolutional layers exploit the fact that an interesting pattern can occur in any region of the image, and regions are contiguous blocks of pixels. But one of the reasons why researchers are excited about deep learning is the potential for the model to learn useful features from raw data. Now, convolutional neural networks can extract informative features from images, eliminating the need of traditional manual image processing methods.
Example with R Code
Example with R Code
Data Source: Data file for CNN
R Code:
# Convolutional Neural Networks# Load packageslibrary(keras)library(EBImage)
# Read Imagessetwd('---')pic1 <- c('p1.jpg', 'p2.jpg', 'p3.jpg', 'p4.jpg', 'p5.jpg', 'c1.jpg', 'c2.jpg', 'c3.jpg', 'c4.jpg', 'c5.jpg', 'b1.jpg', 'b2.jpg', 'b3.jpg', 'b4.jpg', 'b5.jpg')train <- list()for (i in 1:15) {train[[i]] <- readImage(pic1[i])}
pic2 <- c('p6.jpg', 'c6.jpg', 'b6.jpg')test <- list()for (i in 1:3) {test[[i]] <- readImage(pic2[i])}
# Exploreprint(train[[12]])summary(train[[12]])display(train[[12]])plot(train[[12]])
par(mfrow = c(---, ---))for (i in 1:15) plot(train[[i]])par(mfrow = c(1,1))
# Resize & combinestr(train)for (i in 1:15) {train[[i]] <- resize(train[[i]], 100, 100)}for (i in 1:3) {test[[i]] <- ---(test[[i]], 100, 100)}
train <- combine(train)x <- tile(train, 5)display(x, title='Pictures')
test <- ---(test)y <- ---(test, 3)---(y, title = 'Pics')
# Reorder dimensiontrain <- ---(train, c(4, 1, 2, 3))test <- ---(test, c(4, 1, 2, 3))str(train)
# Responsetrainy <- c(0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2)testy <- c(0, 1, 2)
# One hot encodingtrainLabels <- ---(trainy)testLabels <- ---(testy)
# Modelmodel <- keras_model_sequential()
model %>% layer_conv_2d(filters = 32, kernel_size = c(3,3), activation = '---', input_shape = c(---, ---, ---)) %>% ---(filters = 32, kernel_size = c(3,3), activation = 'relu') %>% ---(pool_size = c(2,2)) %>% ---(rate = 0.25) %>% layer_conv_2d(filters = 64, kernel_size = c(3,3), activation = 'relu') %>% layer_conv_2d(filters = 64, kernel_size = c(3,3), activation = 'relu') %>% layer_max_pooling_2d(pool_size = c(2,2)) %>% layer_dropout(rate = 0.25) %>% ---() %>% layer_dense(units = 256, activation = 'relu') %>% layer_dropout(rate=0.25) %>% layer_dense(units = 3, activation = '---') %>% compile(loss = '---', optimizer = optimizer_sgd(lr = 0.01, decay = 1e-6, momentum = 0.9, nesterov = T), metrics = c('---'))summary(model)
# Fit modelhistory <- model %>% fit(train, trainLabels, --- = 60, --- = 32, validation_split = 0.2, validation_data = list(test, testLabels))plot(history)
# Evaluation & Prediction - train datamodel %>% ---(train, trainLabels)pred <- model %>% ---(train)table(Predicted = pred, Actual = trainy)
prob <- model %>% ---(train)cbind(prob, Predicted_class = pred, Actual = trainy)
# Evaluation & Prediction - test datamodel %>% ---(test, testLabels)pred <- model %>% ---(test)table(Predicted = pred, Actual = testy)
prob <- model %>% ---(test)---(prob, Predicted_class = pred, Actual = testy)
Useful and Important Sources
Useful and Important Sources
Page updated
Google Sites
Report abuse