Review on Hand-Drawn Diagram Recognition using Convolutional Neural Network

Abstract

People have a remarkable capacity to recognize freehand sketch drawings, even though these drawings have only the most fundamental theoretical and structural underpinnings. Understanding freehand drawings may be challenging when using automated approaches due to the abstract nature of their compositions and the variety of styles they include. In the course of working on this project, we want to make use of convolutional neural networks in the hopes of developing a freehand sketch recognition method that is both quick and accurate. To be more precise, we want to develop a Keras model that can classify sketches using Google's 'Quick, Draw!' dataset. This dataset comprises more than 50 million drawings that have been separated into 345 different categorization groups. This kind of technology will be very helpful in a wide range of applications, such as human-computer interaction, sketch-based search, game creation, and teaching, to name just a few of them. Because of the COVID-19 epidemic, everything, including education and commerce, has shifted to being conducted online. For online instruction, a stylus or pen is required to create diagrams that may later be altered on a computer. It was difficult to recognize the diagrams since they were so flexible and included a great deal of information. In addition, different users' sketching styles may be distinguished from one another. To proceed with the recognition process, it is important to ascertain whether or not the document in question contains a diagram. In this section, we will take a look at some of the many ways that may be used when identifying hand-drawn diagrams. A cursory investigation of mode identification strategies is also carried out here.