Recognizing chemical structures drawn by hand using deep learning algorithms and predict probable chemical structure

Chemists frequently use structure diagrams created by hand to express concepts regarding organic molecules. However, the simplicity of use, naturalness, and speed of drawing on paper is not present in the software used today to specify these structures to a computer because it relies on a conventional mouse and keyboard interface. As a result, we created a sketch-based system that can decipher manually drawn organic chemical diagrams. We are transforming a chemical molecule's graphical representation into its typical structural representation. The chemical structure recognition method should identify the graph's nodes and groups and the correct bond labels for each vertex. We offer an approach that builds on cutting-edge techniques to address the issue in the face of the additional challenges posed by hand-drawn molecules, allowing users to sketch out molecules on paper and upload a scanned picture of that drawing to recognize the chemicals. We apply fundamental corner detection techniques to determine the atoms and groups that make up the nodes of the chemical structure graph. We conclude that our corner detection method may be more effective than the commonly used line vectorization algorithms. The critical distinction in our policy is employing a neural network that operates supervised machine learning to categorize bonds according to numerous feature descriptors of bond cross-sections.