Deep Learning

Deep learning refers to several layers through which data is transferred and is an element of machine learning processes based on artificial neural networks. Deep learning architectures have been applied to speech recognition, computer vision, audio recognition, natural language processing, social network filtering, bioinformatics, drug design, machine translation and medical image analysis. Deep Learning Architectures can be facilitated through the following platforms: AlexNet is the first type of deep architecture and is usually first applied to deep neural networks for all computer vision and speech recognition tasks. The pyramidal shape characterises the VGG network; the bottom layers are closer to the image and wide-ranging compared to the apex, which contains deep top layers. GoogleNet – sometimes referred to as Inception Network – its architecture goes even deeper and contains 22 layers compared to 19 layers in VGG.  The other architectures are Residual Networks (ResNet) which are made up of multiple subsequent residual modules and are the basic building blocks of ResNet architecture. In addition, ResNeXt is a state-of-the-art technique for object recognition. Region-Based CNN (Convolutional Neural Network) architecture is the most influential of all the deep learning architectures and has been applied to object detection problems. YOLO (You Only Look Once) acts as a real-time system built on deep learning for solving image detection problems processing up to 40 images in a second.  Furthermore, squeezeNet architecture is useful in low bandwidth scenarios like mobile platforms. SegNet is a deep learning architecture applied to solve image segmentation problems. It consists of a sequence of processing layers called encoders followed by a corresponding set of decoders for pixel-wise classification. GAN (Generative Adversarial Network) is a artificial neural networks to generate an entirely new image which is not present in the training dataset but is accurate enough to be in the dataset