Image Transformer

基于Transformer的图像生成自回归模型.

Through-Wall Pose Imaging in Real-Time with a Many-to-Many Encoder/Decoder Paradigm

通过射频信号重建视频中的人体姿态.

Expressive Body Capture: 3D Hands, Face, and Body from a Single Image

SMPLify-X：从单张图像重建3D人体、手部和表情.

Batch Normalization Biases Residual Blocks Towards the Identity Function in Deep Networks

批归一化使深度网络中的残差块偏向于恒等函数.

Bottleneck Transformers for Visual Recognition

BotNet：CNN与Transformer结合的backbone.

SA-Net: Shuffle Attention for Deep Convolutional Neural Networks

SANet：通过特征分组和通道置换实现轻量型置换注意力.