Some Collections Around MMSegmentation
本文收集一系列 MMSegmentation收录的网络结构.
以Res50-D8 在cityscapes上的性能
| Methods | mIoU | FPS |
|---|---|---|
| DeepLabV3+ | 79.61 | 3.94 |
| DNLNet | 78.61 | 2.56 |
| NonLocal | 78.24 | 2.72 |
| PSPNet | 77.85 | 4.07 |
| GCNet | 77.69 | 3.93 |
| ANN | 77.40 | 3.71 |
| OCRNet | 74.30 | 10.45 |
| FCN | 72.25 | 4.17 |
Pyramid Scene Parsing Network (PSPNetwork)
Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation

GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond

Asymmetric Non-local Neural Networks for Semantic Segmentation (ANN)

Object-Contextual Representations for Semantic Segmentation

Disentangled Non-Local Neural Networks (DNL)

图中whiten的操作指的是减去关于channel通道的均值.
