'Per-Pixel Classification is Not All You Need for Semantic Segmentation'
1. Threoretical Background
$$\mathcal{F} \in \mathbb{R}^{C_{\mathcal{F}} \times \frac{H}{S} \times \frac{W}{S}}$$
$${p_{i} \in \Delta^{K + 1}}^{N}{i = 1}$$
$$\epsilon{\text{mask}} \in \mathbb{R}^{C_{\epsilon} \times N}$$