Vision Transformer

Vision Transformers (ViTs) adapt the transformer architecture, initially designed for natural language processing, to image analysis by treating images as sequences of patches. Current research focuses on improving ViT efficiency and robustness through techniques like token pruning, attention engineering, and hybrid models combining ViTs with convolutional neural networks or other architectures (e.g., Mamba). These advancements are driving progress in various applications, including medical image analysis, object detection, and spatiotemporal prediction, by offering improved accuracy and efficiency compared to traditional convolutional neural networks in specific tasks.

Papers

October 19, 2023

October 17, 2023

USDC: Unified Static and Dynamic Compression for Visual Transformer
Huan Yuan, Chao Liao, Jianchao Tan, Peng Yao, Jiyuan Jia, Bin Chen, Chengru Song, Di Zhang
Vision Transformer Unifying Theory Visual Transformer Vision Transformer Model Accurate Training Transformer Baseline

October 16, 2023

Interpreting and Controlling Vision Foundation Models via Text Explanations
Haozhe Chen, Junfeng Yang, Carl Vondrick, Chengzhi Mao
Vision Transformer Foundation Model Vision Foundation Model Machine Translated Textual Explanation Latent Token Reasoning Behavior

October 15, 2023

October 13, 2023

October 12, 2023

LEMON: Lossless model expansion
Yite Wang, Jiahao Su, Hanlin Lu, Cong Xie, Tianyi Liu, Jianbo Yuan, Haibin Lin, Ruoyu Sun, Hongxia Yang
Deep Neural Network Vision Transformer Transformer Megatron Decepticons Large Model Model Training Model Expansion

October 11, 2023

October 10, 2023

October 9, 2023

Vision Transformer

Papers

2D-3D Interlaced Transformer for Point Cloud Segmentation with Scene-Level Supervision

Model Merging by Uncertainty-Based Gradient Matching

USDC: Unified Static and Dynamic Compression for Visual Transformer

Interpreting and Controlling Vision Foundation Models via Text Explanations

Top-K Pooling with Patch Contrastive Learning for Weakly-Supervised Semantic Segmentation

MoEmo Vision Transformer: Integrating Cross-Attention and Movement Vectors in 3D Pose Estimation for HRI Emotion Detection

Tackling Heterogeneity in Medical Federated learning via Vision Transformers

PaLI-3 Vision Language Models: Smaller, Faster, Stronger

Vision Transformers increase efficiency of 3D cardiac CT multi-label segmentation

LEMON: Lossless model expansion

3D TransUNet: Advancing Medical Image Segmentation through Vision Transformers

Accelerating Vision Transformers Based on Heterogeneous Attention Patterns

EViT: An Eagle Vision Transformer with Bi-Fovea Self-Attention

Efficient Adaptation of Large Vision Transformer via Adapter Re-Composing

ViTs are Everywhere: A Comprehensive Study Showcasing Vision Transformers in Different Domain

No Token Left Behind: Efficient Vision Transformer via Dynamic Token Idling

Plug n' Play: Channel Shuffle Module for Enhancing Tiny Vision Transformers

A Simple and Robust Framework for Cross-Modality Medical Image Segmentation applied to Vision Transformers

RetSeg: Retention-based Colorectal Polyps Segmentation Network

Hierarchical Side-Tuning for Vision Transformers