The paper introduces EdgeConv, a novel neural network module designed for processing point clouds using convolutional neural networks (CNNs). Point clouds, which are collections of 3D points, are widely used in computer graphics and 3D data acquisition. While traditional methods often rely on handcrafted features, the authors propose EdgeConv to learn local geometric features directly from point clouds. EdgeConv operates on dynamically computed graphs at each layer, capturing both local and global shape properties. The module is differentiable and can be integrated into existing CNN architectures. The authors demonstrate the effectiveness of EdgeConv on various benchmarks, including ModelNet40, ShapeNetPart, and S3DIS, achieving state-of-the-art performance in classification and segmentation tasks. Key contributions include the introduction of EdgeConv, its integration into PointNet, and extensive experimental validation. The paper also discusses related work and provides a detailed evaluation of EdgeConv's performance and complexity.The paper introduces EdgeConv, a novel neural network module designed for processing point clouds using convolutional neural networks (CNNs). Point clouds, which are collections of 3D points, are widely used in computer graphics and 3D data acquisition. While traditional methods often rely on handcrafted features, the authors propose EdgeConv to learn local geometric features directly from point clouds. EdgeConv operates on dynamically computed graphs at each layer, capturing both local and global shape properties. The module is differentiable and can be integrated into existing CNN architectures. The authors demonstrate the effectiveness of EdgeConv on various benchmarks, including ModelNet40, ShapeNetPart, and S3DIS, achieving state-of-the-art performance in classification and segmentation tasks. Key contributions include the introduction of EdgeConv, its integration into PointNet, and extensive experimental validation. The paper also discusses related work and provides a detailed evaluation of EdgeConv's performance and complexity.