ProjectsProject Details

The project aims to develop an innovative application that enhances the learning experience in physics classrooms. The objective is to create an application capable of transforming a photo of an illustration on a blackboard into a dynamic simulation of the depicted physical system. This project combines the fields of deep learning, image processing, physical modeling, and programming, building upon the achievements of a previous project that laid the groundwork for object detection, classification, localization, mapping, and animation of the system. In the current project, we focused on improving the accuracy of object mapping, and on integrating all system modules into a cohesive application. In the initial stage of the project, we improved the synthetic image generator from the previous project to incorporate object mapping. This enhanced generator enabled us to generate a large image dataset, which we used to train a simple deep neural network. To improve mapping results, we refined the dataset. Finally, the newly developed mapping module was integrated with the existing modules from the previous project. The result is a system that operates on the users smartphone, employing a client-server architecture.