Image classification using CNN

Image classification using CNN

• 1.
  Image Classification Using CNNand Tensorflow
• 2.
  Team members ➔ NouraHussein Fekry @noura_7ussein ➔ Mohamed Mahmoud Shehata @Shehata
• 3.
  Project Description Building amodel to classify images into their appropriate categories using the CIFAR-10 dataset. Using CNN (python 3.6 and tensorflow) ● TensorFlow is an open-source software library for dataflow programming across a range of tasks. It is a symbolic math library, and is also used for machine learning applications such as neural networks.
• 4.
  The CIFAR-10 Dataset TheCIFAR-10 dataset contains 60 thousand images of various objects like: Airplanes, Cars, Birds, trucks, deers, dogs, frogs, horses, cats and ships. The Images are (32*32*3) It contains 50 thousand images as a training data and 10 thousand testing data.
• 5.
  Phases Preprocess Data Phase CNN Phase Gettingthe data Preprocessing functions Building the network Training the network Model Testing
• 6.
  Preprocess functions ● Normalization Transformingimage X into a NumPy array. ● One Hot Encoded Labeling the images into 0 or 1, by indexing the classes, giving 1 to the class and 0 to the rest of the classes. ● Randomize the data Making sure that the data is as random as possible, to insure maximum randomization
• 7.
  So, why didwe use a CNN? In machine learning, a convolutional neural network is a class of deep, feed-forward artificial neural networks that has successfully been applied to analyzing visual imagery. CNNs use a variation of multilayer perceptrons designed to require minimal preprocessing. They are also known as shift invariant or space invariant artificial neural networks (SIANN), based on their shared-weights architecture and translation invariance characteristics.
• 8.
  CNN Building Blocks ●Convolutional Layer ● Pooling Layer ● ReLU Layer ● Fully Connected Layer ● Loss Layer
• 9.
  CNN’s design consistsof Convolutional, pooling, fully connected and weights.
• 10.
  Building the network ➔Input ➔ Convolution and Max Pooling layer ➔ Flatten Layer ➔ Fully Connected Layer ➔ Output Layer
• 11.
  Training the neural net ●Single Optimization ● Hyperparameters ● Training on a single CIFAR-10 Batch ● Fully Trained Model Testing Model ● Accuracy: 71.44% ● Softmax Predictions
• 12.
  Test Samples
• 13.
  Resources: ● https://en.wikipedia.org/wiki/Convolu tional_neural_network ● https://www.youtube.com/watch?v=F Tr3n7uBIuE
• 14.
  Thank you.