Demystifying a Classical Image Classifier and Leveraging its Accuracy

Deep learning can be called as a type of machine learning which is concerned with computer performing activities that are generally performed by humans. Deep learning finds its use behind driverless cars, helping them understand and distinguish a stop sign, or to tell apart a pedestrian from another car. It is highly used in IoT (Internet of Things) like in Television, speaker, kitchen appliances and many more. Deep learning is receiving a lot of attention lately and for good reason. It’s achieving results that were not possible before. In deep learning, the learning process takes place from the information gathered from images, audio and other sources and then the computational model performs the classification. Deep learning models can achieve the accuracy, which sometimes exceeding human-level performance. In this, initially, a model is trained by using a labelled data, which is mostly verified and neural network architectures that mainly consist many layers. In this paper we mainly focus on the optimization of different parameters of convolutional neural network of deep learning for classifying 8000 labelled natural images of cat and dog. Various level of optimization have been done to improve the performance level of the network and finally, we achieved the best classification accuracy of 93.10% achieved the best classification accuracy of 93.10%.