Face Generation using Deep Convolutional Generative Adversarial Neural Network

Due to the huge availability of data, it is difficult to classify/process images at a higher speed and accuracy. The first technique was in the field of computer vision and it used image data for face recognition and detection of an object from the image but later Convolutional Neural Networks (CNNs) took place. CNNs are used for feature detections by looking at the image and try to check if certain features are present in the image and then it classifies the image accordingly. Acquiring and processing the dataset for the Machine learning technique is one of the time-consuming processes, so Generative Adversarial Neural Network ( GAN) are introduced. GAN typically work with image dataset but they are difficult to train. This paper explores the potential of GAN to generate realistic images. Deep Convolutional Generative Adversarial Networks (DCGAN) is used to generate new images that are not in the dataset. DCGAN has a great success in generating the new images. MNIST (Modified National Institute of Standards and Technology dataset) contains images of handwritten digits dataset and CelebA dataset contains images of celebrities are used, performing the adversarial learning on it and try to generate new images as same as the MNIST and CelebA datasets.