A Detailed Survey on Deep Learning Techniques for Real-Time Image Classification, Recognition and Analysis

Abstract

The portion of machine learning (ML) is deep learning (DL). Machine learning (ML) is the study of computer algorithms. It constructs a model using training data, often referred to as sample data for prediction. Artificial intelligence (AI) is a sub-branch in the field of computer science (CS). With the help of training data, ML algorithms construct a model, often referred to as sample data for prediction and decision-making. Programming is often needed to do something with computers, but by implementing a model generated by machine learning algorithms it can prevent programming and to do what programming can do without programming assistance. Machine learning algorithms can be used widely in various real-world applications such as e-mail filtering, computer networks, natural language processing, search engines, telecommunications, Internet fraud detection and DNA sequence classification. Three types of learning algorithms are present: supervised, unsupervised and reinforcement. Machine learning ML is a widely used multidisciplinary field which uses various training models and algorithms to predict, classify and analyse any statistical data by the use of computer science algorithms. This paper is going to address deep learning techniques such as single-shot detector (SSD), scale-invariant feature transform (sfit), histogram of oriented gradient (HOG) and many more. The main aim is to detect cybercrimes through the assistance of the above-mentioned techniques.