Table 1.

mIoU (%) of various augmentations with the four segmentation models where $N_{train}=4$ and $N_{test}=all$.

Augmentation	$\mathbf{UNet}$	$\mathbf{FPN}$	$\mathbf{DeepLabV}3+$	$\mathbf{HRNet}-\mathbf{OCR}$
CutMix [24] (2019)	89.19	84.98	85.52	79.35
Copy-Paste [25] (2021)	89.19	86.65	83.64	84.77
Ours	90.77	88.45	88.41	87.28