Pooling Revisited: Your Receptive Field is Suboptimal
DynOPool: 学习最优感受野的动态优化池化.
A ConvNet for the 2020s
ConvNeXt: 设计现代化卷积神经网络.
Deformable Convolutional Networks
可变形卷积神经网络.
Deformable ConvNets v2: More Deformable, Better Results
改进的可变形卷积神经网络.
CondConv: Conditionally Parameterized Convolutions for Efficient Inference
CondConv：用于高效推理的条件参数化卷积.
Dynamic Convolution: Attention over Convolution Kernels
DynamicConv：卷积核上的注意力.