Look into Person: Self-supervised Structure-sensitive Learning and A New Benchmark for Human Parsing

Benchmark Model Rank Results
semantic-segmentation-on-lip-valAttention+SSL (ResNet-101)#12mIoU: 44.73%