Dr. Josep Alemany, a researcher at the Universidad Internacional Iberoamericana (International Iberoamerican University, UNIB), is participating in a study that proposes a revolutionary model that combines neural networks and classification algorithms to overcome current challenges in the recognition of hand-drawn geometric figures.
Hand-drawn geometric figures, such as circles, squares and triangles, are a quick and natural way of expressing mathematical concepts. However, these representations are often imperfect, with irregular lines, variable angles and incomplete shapes. Traditionally, pattern recognition systems have faced difficulties in interpreting these images accurately, limiting their application in educational technology tools, mathematical tools and digital note-taking applications, where correct interpretation can significantly improve the user experience. Therefore, there is an urgent need for more robust and adaptable methods that can overcome the limitations of traditional approaches.
To this end, the study introduces a model called CnN-RFc, which combines convolutional neural networks (CNN) to extract spatial characteristics from images, with a random forest classifier (RFC) that optimizes the interpretation of these characteristics. This hybrid approach allowed for greater accuracy in the recognition of geometric figures, even when they are drawn irregularly or in different orientations.
In addition, the performance of five advanced machine learning algorithms was evaluated, applying hyperparameter optimization and cross-validation techniques to improve model generalization. The results show that the proposed approach outperformed existing methods, achieving an outstanding accuracy of 98%, setting a new standard in the field.
The model has significant practical applications in educational platforms and computer-aided design (CAD) tools. In education, it allows students to draw geometric shapes by hand, which the software can recognize and interpret, offering instant feedback and fostering interactive learning. In the case of CAD tools, the model streamlines design by accurately digitizing preliminary sketches, improving accessibility and efficiency for professionals and beginners alike. In addition, strategies have been proposed to implement the model on mobile devices, reducing its size from 15.51 MB to a range of 5-10 MB through compression, lightweight architectures and on-demand calculations, ensuring efficient use of memory without compromising accuracy.
However, the model faces certain limitations. Its performance could be affected in more complex data sets that include shapes with intricate details, overlapping lines or varied drawing styles. These variations, such as line thicknesses or artistic styles, could influence its accuracy, as deep learning models often have difficulty generalizing with data that deviates from the training distribution.
In the future, the model is planned to be extended to more complex categories, such as flowcharts or handwritten texts, and combined with other techniques to further improve its accuracy and robustness. This advance promises to contribute significantly to the field of image recognition and the development of interactive tools in various sectors.
In conclusion, this study represents a significant advance in the recognition of hand-drawn geometric figures, with implications that go beyond the academic sphere. By combining advanced machine learning techniques, this study has opened the door to a new era of human-machine interaction, where creativity and technology converge in a harmonious way.
If you want to know more about this study, click here.
To read more research, see the repository of UNIB.
The International Iberoamerican University (UNIB) offers a Master's Degree in Strategic Management with a Specialization in Information Technology. A program that provides professionals with the skills and abilities necessary to manage executive positions and direct organizational change projects using ICTs or to advise companies that want to join the competitiveness of new businesses. Join the business development by studying our master's program and get ready to take your career to the next level!