语义分割 DeepLab(v1和v2)
DeepLab v1:Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs ——ICLR 2015
论文地址:https://arxiv.org/pdf/1412.7062v3.pdf
DeepLab v2:DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs ——ICLR 2017
论文地址:https://arxiv.org/pdf/1606.00915.pdf
主要贡献:
1. 使用了膨胀卷积;
2. 提出了在空间维度上实现金字塔型的空洞池化atrous spatial pyramid pooling(ASPP);
3. 使用了全连接条件随机场(CRF)。