FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence

Code

• github.com/google-research/fixmatch
  OfficialIn papertf★ 1,208
• github.com/kekmodel/FixMatch-pytorch
  pytorch★ 797
• github.com/LeeDoYup/FixMatch-pytorch
  pytorch★ 204
• github.com/amazon-research/exponential-moving-average-normalization
  pytorch★ 105
• github.com/google-research/crest
  tf★ 100
• github.com/KaiyangZhou/ssdg-benchmark
  pytorch★ 72
• github.com/valencebond/FixMatch_pytorch
  pytorch★ 48
• github.com/CoinCheung/fixmatch
  pytorch★ 36
• github.com/CoinCheung/fixmatch-pytorch
  pytorch★ 36
• github.com/gomezzz/MSMatch
  pytorch★ 29

Abstract

Semi-supervised learning (SSL) provides an effective means of leveraging unlabeled data to improve a model's performance. In this paper, we demonstrate the power of a simple combination of two common SSL methods: consistency regularization and pseudo-labeling. Our algorithm, FixMatch, first generates pseudo-labels using the model's predictions on weakly-augmented unlabeled images. For a given image, the pseudo-label is only retained if the model produces a high-confidence prediction. The model is then trained to predict the pseudo-label when fed a strongly-augmented version of the same image. Despite its simplicity, we show that FixMatch achieves state-of-the-art performance across a variety of standard semi-supervised learning benchmarks, including 94.93% accuracy on CIFAR-10 with 250 labels and 88.61% accuracy with 40 -- just 4 labels per class. Since FixMatch bears many similarities to existing SSL methods that achieve worse performance, we carry out an extensive ablation study to tease apart the experimental factors that are most important to FixMatch's success. We make our code available at https://github.com/google-research/fixmatch.

Tasks

Benchmark Results

Reproductions