Towards Better Accuracy-efficiency Trade-offs: Divide and Co-training

Open paper
Benchmark Model Rank Results
image-classification-on-cifar-10PyramidNet-272, S=4#42Percentage correct: 98.71
image-classification-on-cifar-10WRN-40-10, S=4#56Percentage correct: 98.38
image-classification-on-cifar-10WRN-28-10, S=4#57Percentage correct: 98.32
image-classification-on-cifar-10Shake-Shake 26 2x96d, S=4#58Percentage correct: 98.31
image-classification-on-cifar-100PyramidNet-272, S=4#51Percentage correct: 89.46PARAMS: 32.8M
image-classification-on-cifar-100DenseNet-BC-190, S=4#69Percentage correct: 87.44PARAMS: 26.3M
image-classification-on-cifar-100WRN-40-10, S=4#72Percentage correct: 86.90
image-classification-on-cifar-100WRN-28-10, S=4#84Percentage correct: 85.74
image-classification-on-imagenetSE-ResNeXt-101, 64x4d, S=2(320px)#439Top 1 Accuracy: 83.6%Number of params: 98MGFLOPs: 38.2
image-classification-on-imagenetSE-ResNeXt-101, 64x4d, S=2(416px)#466Top 1 Accuracy: 83.34%Number of params: 98MGFLOPs: 61.1
image-classification-on-imagenetResNeXt-101, 64x4d, S=2(224px)#608Top 1 Accuracy: 82.13%Number of params: 88.6MGFLOPs: 18.8

Rank counts only results with a code link.