Image Classification on ImageNet (val) (Accuracy, Throughput, Latency)
78.1AccuracyTrivialAugment
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TrivialAugmentSearch Overhead=0x, Backbone=ResNet2021.03 | 78.1 | — | — | |
| AutoAugmentSearch Overhead=40-800x, Backbone=ResNet2021.03 | 77.6 | — | — | |
| RandAugmentSearch Overhead=4-80x, Backbone=ResNet2021.03 | 77.6 | — | — | |
| Fast AutoAugmentSearch Overhead=1x, Backbone=ResNet2021.03 | 77.6 | — | — | |
| ChaseModel=ResNet-50, Sp=0.9, Sc=0.42023.05 | 74.03 | 71.54 | 30.13 | |
| ChaseModel=ResNet-50, Sp=0.9, Sc=0.52023.05 | 73.39 | 78.19 | 27.55 | |
| Chase (prune skip)Model=ResNet-50, Sp=0.9, Sc=0.42023.05 | 73.24 | 87.38 | 24.53 | |
| RigLModel=ResNet-50, Sp=0.9, Sc=0.02023.05 | 73 | 59.5 | 35.79 | |
| Chase (prune skip)Model=ResNet-50, Sp=0.9, Sc=0.52023.05 | 72.71 | 99.97 | 21.55 |