International Journal of Scientific Research and Engineering Development-– Volume 4 Issue 1, Jan-Feb 2021 Available at www.ijsred.com ISSN : 2581-7175 ©IJSRED: All Rights are Reserved Page 302 Image Segmentation and Recognition Based on Deep Learning David Taylor 1 ,Emily Smith 1 1.LimeStoneUniversity, South Carolina,USA Abstract:Image segmentation these days have gained lot of interestfor the researchers of computer vision and machine learning. Deep Learning techniques have achieved accuracy in many fields like medicine, automobiles- self driving, indoor navigation etc., We discuss first about what is deep learning and how it has evolved, and how it is helpful in different fields and convolutional neural networks, architectures, and their related calculations of accuracy. We also discuss about some research paper and their achieved accuracy in different fields of medicine diagnosis using Deep learning techniques. Introduction: Deep learning is subset of machine learning, and machine learning is a subset of artificial intelligence. Artificial intelligence is a technique which will let computers behave like humans. Machine learning is a set of algorithms that is trained on data from different sources and will enable computers behave like this. Deep learning is also a form of machine learning which is taken as inspiration from human brain. Deep learning continuously analyzes data with the given logics and inorder to complete this process deep learning used a set of algorithms called neural networks.Neural network’s design is based on human brain structure. Our brain will identify different objects and different information same way neural networks Convolutional Neural networks: Visual systems structure is taken into inspiration for convolutional Neural Networks (CNNs). CNN is a kind of artificial neural network and are proved immensely powerful [20]. Every ANN will have three layers input layer, hidden layer and output layer. To the input layer image is fed as input and the hidden layer is used for training the architecture based on input images that are fed, and the output layer gives the output. CNNs will have two components, out of which one component will extract the features and other component will do the classification [21]. ANN learn by using weights, by adjusting the weights ANN will decide what to pass to the next signal. While training the ANN we can control the weights. Activation function is also applied to the neural network and it will decide based on the conditions whether to pass the signal along or not to the next layer. There are different models of CNNs. Based on local connectivity which is between neurons and on image transformations in a computational model are found in Neural networks. This specifies that Neurons which has same parameters are applied on previous layer patches at multiple locations. CNN contains three neural layers. • Convolutional Layers • Pooling Layers • Fully connected layers Each Layer is used for different purposes. Each Layer of CNN takes the input and gives output which will be input for next layer, which will have fully connected layers. CNNs are successfully used these days successfully in vision applications using computer vision applications, like image recognition, and detecting objects, robotics, and in self driving cars. CNNs have been extensively used in medicine for skin cancer classification [9]. RESEARCH ARTICLE OPEN ACCESS