Dataset	PASCAL VOC 2012	train: 13, 000 images class: 20 object classes and 1 background class labeled set: 1/8
Metric	mIoU
Backbone	ResNet-50	Pre-trained on ImageNet
Segmentation head	DeepLabv3+	Randomly
Optimizer	SGD
GPU num	4

IMG	MT	CCT	CPS	CPS+CUTMIX

Method	mIoU	Pixel_ACC
MT	70.575	93.265
CCT	70.650	93.168
GCT	70.470	93.023
CPS	73.201	94.028
CPS+CutMix	73.971	94.160

• Deep Variational Information Bottleneck 2017 ICLR VIB
• ON THE INFORMATION BOTTLENECK THEORY OF DEEP LEARNING 2017 ICLR
• Adversarial Feature Distribution Alignment for Semi-Supervised Learning
• Distilling Robust and Non-Robust Features in Adversarial Examples by Information Bottleneck
• VARIATIONAL DISCRIMINATOR BOTTLENECK: IMPROVING IMITATION LEARNING, INVERSE RL, AND GANS BY CONSTRAINING INFORMATION FLOW