The GIST and RIST of Iterative Self-Training for Semi-Supervised Segmentation

31 Mar 2021  ·  Eu Wern Teh, Terrance DeVries, Brendan Duke, Ruowei Jiang, Parham Aarabi, Graham W. Taylor ·

We consider the task of semi-supervised semantic segmentation, where we aim to produce pixel-wise semantic object masks given only a small number of human-labeled training examples. We focus on iterative self-training methods in which we explore the behavior of self-training over multiple refinement stages... We show that iterative self-training leads to performance degradation if done na\"ively with a fixed ratio of human-labeled to pseudo-labeled training examples. We propose Greedy Iterative Self-Training (GIST) and Random Iterative Self-Training (RIST) strategies that alternate between training on either human-labeled data or pseudo-labeled data at each refinement stage, resulting in a performance boost rather than degradation. We further show that GIST and RIST can be combined with existing semi-supervised learning methods to boost performance. read more

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Semi-Supervised Semantic Segmentation Cityscapes 100 samples labeled GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) Validation mIoU 58.70% # 4
Semi-Supervised Semantic Segmentation Cityscapes 12.5% labeled GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) Validation mIoU 62.57% # 8
Semi-Supervised Semantic Segmentation Cityscapes 25% labeled GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) Validation mIoU 65.14% # 7
Semi-Supervised Semantic Segmentation Cityscapes 2% labeled GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) Validation mIoU 53.51% # 1
Semi-Supervised Semantic Segmentation Cityscapes 5% labeled GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) Validation mIoU 59.98% # 1
Semi-Supervised Semantic Segmentation Pascal VOC 2012 12.5% labeled GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) Validation mIoU 70.76% # 11
Semi-Supervised Semantic Segmentation Pascal VOC 2012 2% labeled GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) Validation mIoU 67.21% # 3
Semi-Supervised Semantic Segmentation Pascal VOC 2012 5% labeled GIST and RIST (DeepLabv2 with ResNet101, MSCOCO pre-trained) Validation mIoU 69.40% # 5

Methods


No methods listed for this paper. Add relevant methods here