MUMBAI, India, July 30 -- Intellectual Property India has published a patent application (202641088017 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on July 18, 2026, for A Graph-Based Convolutional Neural Network System For Efficient 3d Object Detection From Point Cloud Data.
Inventor includes Dr. G. Ramesh Chandra.
The application for the patent was published on July 24, 2026, under issue no. 30/2026.
Abstract: ABSTRACT [0022] Traditional 3D object detection methods, such as voxelization and point-wise processing, often face challenges due to the irregular and sparse nature of point cloud data. These techniques either involve hand-crafted features that lack adaptability or grid-based approaches that are computationally expensive and prone to information loss. As a result, they struggle to fully capture the complexity of real-world 3D environments, affecting both the accuracy and efficiency of object detection. To address these limitations, we propose a graph-based Convolutional Neural Network (CNN) approach for 3D object detection using point clouds from the KITTI dataset. In this method, raw point clouds are first converted into mesh structures, where vertices represent individual points and edges capture their geometric relationships. This transformation allows for the representation of both local and global structural information within the scene. [0023] Graph-based CNNs are then applied to these mesh structures, enabling the model to learn features more effectively by leveraging the connectivity and spatial context of the data. This not only enhances detection accuracy but also improves computational efficiency by avoiding the drawbacks of dense voxelization and high-dimensional input representations. Our method offers a scalable and flexible solution for real-time applications, making it well-suited for autonomous driving and robotics. By overcoming the constraints of traditional techniques, this graph-based approach provides a robust framework for accurate and efficient 3D object detection in complex environments.
Disclaimer: Curated by HT Syndication.
