Geometry Biased Transformer

Geometry-biased transformers are a novel class of neural network architectures designed to improve the accuracy and robustness of various computer vision tasks by explicitly incorporating geometric information into the transformer's attention mechanism. Current research focuses on applying these architectures to problems like 3D human pose estimation, point cloud generation and processing, and novel view synthesis, often employing specialized attention mechanisms that leverage spatial relationships between data points or views. This approach addresses limitations of traditional methods, particularly in scenarios with occlusions, limited data, or complex geometries, leading to more accurate and reliable results in applications ranging from autonomous driving to medical image analysis.

Papers

July 29, 2024

Text2LiDAR: Text-guided LiDAR Point Cloud Generation via Equirectangular Transformer
Yang Wu, Kaihua Zhang, Jianjun Qian, Jin Xie, Jian Yang
LiDAR Point Cloud LiDAR Data Lidar Feature Geometry Biased Transformer

December 28, 2023

Geometry-Biased Transformer for Robust Multi-View 3D Human Pose Reconstruction
Olivier Moliner, Sangxia Huang, Kalle Åström
Attention Mechanism 3D Pose 3D Human Pose Geometry Biased Transformer

September 23, 2023

RBFormer: Improve Adversarial Robustness of Transformer by Robust Bias
Hao Cheng, Jinhao Duan, Hui Li, Lyutianyang Zhang, Jiahang Cao, Ping Wang, Jize Zhang, Kaidi Xu, Renjing Xu
Transformer Based Adversarial Robustness Convolutional Structure Transformer Based Structure Robustness Bias Geometry Biased Transformer

September 9, 2023

Few-Shot Medical Image Segmentation via a Region-enhanced Prototypical Transformer
Yazhou Zhu, Shidong Wang, Tong Xin, Haofeng Zhang
Medical Image Segmentation Shot Learning Medical Image Datasets Region Transformer Geometry Biased Transformer

April 27, 2023

Exploiting Inductive Bias in Transformer for Point Cloud Classification and Segmentation
Zihao Li, Pan Gao, Hui Yuan, Ran Wei, Manoranjan Paul
Transformer Based Segmentation Based Approach Inductive Bias Local Feature Point Cloud Classification Locality Sensitive Geometry Biased Transformer

April 16, 2023

EGformer: Equirectangular Geometry-biased Transformer for 360 Depth Estimation
Ilwi Yun, Chanyong Shin, Hyunku Lee, Hyuk-Jae Lee, Chae Eun Rhee
Attention Mechanism Depth Estimation Large Receptive Field Equirectangular Projection Geometry Biased Transformer

January 11, 2023

Geometry-biased Transformers for Novel View Synthesis
Naveen Venkat, Mayank Agarwal, Maneesh Singh, Shubham Tulsiani
Novel View Synthesis Inductive Bias Geometry Biased Transformer