Rock-Scissors-Paper Image Classification using Tensorflow and Keras.
This is my practice in deep learning for image classification using Convolutional Neural Network (CNN) from Keras and TensorFlow library in Python, I am using Go Colaboratory as IDE.
CNN is a class of artificial neural network (ANN), most commonly applied for visual imagery. CNN is neural network for recognize image based on the pixels contained in the image. A more optimal technique is to use a convolutional layer where instead of recognizing object based on pixel, the neural network can recognize image based on attributes that have more information. Convolutional filter can recognize attributes in the object using filter, filter is matrix containing numbers.
Convulsions is process for applying filter to image, in the convulsions process there is a matrix multiplication of the filter and the area of image.
When the convolution process is complete, the results of the convolution can be used as input for an Multi Layer Percepton. In ANN, generally after convolution process on the input image, a pooling process will be carried out. Pooling is the process of reducing the image resolution while retaining the information in the image.
This is the arcitecture of CNN.
Visit my github page or click
to see the code.