Dense FixMatch: a simple semi-supervised learning method for pixel-wise prediction tasks
This work addresses the challenge of reducing labeled data requirements for pixel-wise prediction tasks in computer vision, representing an incremental advancement by adapting an existing method to new domains.
The paper tackled the problem of semi-supervised learning for dense prediction tasks like semantic segmentation by proposing Dense FixMatch, which combines pseudo-labeling and consistency regularization with strong data augmentation. The result showed significant improvements over supervised learning with only labeled data, approaching its performance with one-fourth of the labeled samples on Cityscapes and Pascal VOC datasets.
We propose Dense FixMatch, a simple method for online semi-supervised learning of dense and structured prediction tasks combining pseudo-labeling and consistency regularization via strong data augmentation. We enable the application of FixMatch in semi-supervised learning problems beyond image classification by adding a matching operation on the pseudo-labels. This allows us to still use the full strength of data augmentation pipelines, including geometric transformations. We evaluate it on semi-supervised semantic segmentation on Cityscapes and Pascal VOC with different percentages of labeled data and ablate design choices and hyper-parameters. Dense FixMatch significantly improves results compared to supervised learning using only labeled data, approaching its performance with 1/4 of the labeled samples.