Publications by authors named "Zheng-Ning Liu"
Article Synopsis
- * The paper introduces DGNet, a deep neural network leveraging dual graph pyramids to efficiently handle irregular mesh structures, enhancing feature propagation and local geometric information gathering.
- * Experimental results show that DGNet excels in tasks like shape analysis and scene understanding, outperforming existing methods on benchmarks, with resources available for further exploration.
Beyond Self-Attention: External Attention Using Two Linear Layers for Visual Tasks.
IEEE Trans Pattern Anal Mach Intell
May 2023
Article Synopsis
- Self-attention is crucial for deep learning in visual tasks, but it has a quadratic complexity and doesn't account for correlations between different samples.
- The article introduces "external attention," which uses two small, shared memories and linear layers to replace self-attention, resulting in linear complexity and better handling of sample correlations.
- The new external attention mechanism, combined with a multi-head approach, forms an all-MLP architecture called EAMLP, demonstrating competitive performance in various tasks with lower computational costs.
High-Quality Textured 3D Shape Reconstruction with Cascaded Fully Convolutional Networks.
IEEE Trans Vis Comput Graph
January 2021
We present a learning-based approach to reconstructing high-resolution three-dimensional (3D) shapes with detailed geometry and high-fidelity textures. Albeit extensively studied, algorithms for 3D reconstruction from multi-view depth-and-color (RGB-D) scans are still prone to measurement noise and occlusions; limited scanning or capturing angles also often lead to incomplete reconstructions. Propelled by recent advances in 3D deep learning techniques, in this paper, we introduce a novel computation- and memory-efficient cascaded 3D convolutional network architecture, which learns to reconstruct implicit surface representations as well as the corresponding color information from noisy and imperfect RGB-D maps.
View Article and Find Full Text PDF