Billion-scale semi-supervised learning for image classification

Open paper
Benchmark Model Rank Results
image-classification-on-imagenetResNeXt-101 32x16d (semi-weakly sup.)#276Top 1 Accuracy: 84.8%Number of params: 193M
image-classification-on-imagenetResNeXt-101 32x8d (semi-weakly sup.)#316Top 1 Accuracy: 84.3%Number of params: 88M
image-classification-on-imagenetResNeXt-101 32x4d (semi-weakly sup.)#413Top 1 Accuracy: 83.4%Number of params: 42M
image-classification-on-omnibenchmarkIG-1B#6Average Top-1 Accuracy: 40.4
object-recognition-on-shape-biasSWSL (ResNeXt-101)#9shape bias: 49.8
object-recognition-on-shape-biasSWSL (ResNet-50)#15shape bias: 28.6

Rank counts only results with a code link.