Image Classification on CIFAR-10 (Accuracy and Runtime)
96.5AccuracyPyramidCLIP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PyramidCLIPImage Encoder=ViT-B/16, Pretrain Dataset Size=143M, Evaluation Protocol=Linear probe2022.04 | 96.5 | — | — | |
| CLIP*Image Encoder=ViT-B/16, Pretrain Dataset Size=400M, Evaluation Protocol=Linear probe2022.04 | 96.2 | — | — | |
| PyramidCLIPImage Encoder=ViT-B/32, Pretrain Dataset Size=143M, Evaluation Protocol=Linear probe2022.04 | 96 | — | — | |
| CLIP*Image Encoder=ViT-B/32, Pretrain Dataset Size=400M, Evaluation Protocol=Linear probe2022.04 | 95.1 | — | — | |
| BaselineNetwork=ResNet18, #ReLUs (K)=557.12021.06 | 94.66 | 11.05 | — | |
| Circa (PosZero)Network=ResNet18, #ReLUs (K)=557.1, Stochastic ReLU Mode=PosZero, Truncation Bits=112021.06 | 94.24 | 3.89 | 2.8 | |
| BaselineNetwork=VGG16, #ReLUs (K)=284.72021.06 | 94 | 5.89 | — | |
| Circa (NegPass)Network=ResNet18, #ReLUs (K)=557.1, Stochastic ReLU Mode=NegPass, Truncation Bits=112021.06 | 93.77 | 3.89 | 2.8 | |
| Circa (NegPass)Network=VGG16, #ReLUs (K)=284.7, Stochastic ReLU Mode=NegPass, Truncation Bits=122021.06 | 93.77 | 2.25 | 2.6 | |
| Circa (PosZero)Network=VGG16, #ReLUs (K)=284.7, Stochastic ReLU Mode=PosZero, Truncation Bits=132021.06 | 93.61 | 2.25 | 2.6 | |
| BaselineNetwork=ResNet32, #ReLUs (K)=303.12021.06 | 92.43 | 6.32 | — | |
| Circa (PosZero)Network=ResNet32, #ReLUs (K)=303.1, Stochastic ReLU Mode=PosZero, Truncation Bits=122021.06 | 91.85 | 2.47 | 2.6 | |
| Circa (NegPass)Network=ResNet32, #ReLUs (K)=303.1, Stochastic ReLU Mode=NegPass, Truncation Bits=122021.06 | 91.47 | 2.47 | 2.6 |