Image Classification on ImageNet 2012 (test)
85.7AccuracyEfficientNetV2-L
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EfficientNetV2-LParams=121M, Pre-trained=ImageNet ILSVRC20122021.04 | 85.7 | — | |
| Mahajan et al.Protocol=Pre-trained on non-public Instagram data2018.11 | 85.4 | — | |
| EfficientNetV2-MParams=55M, Pre-trained=ImageNet ILSVRC20122021.04 | 85.1 | — | |
| EfficientNet-B7Params=66M, Pre-trained=ImageNet ILSVRC20122021.04 | 84.7 | — | |
| GPipe (AmoebaNet-B (18, 512))Architecture=AmoebaNet-B (18, 512), Resolution=480x480, Protocol=Pre-training, Crop=Single-crop2018.11 | 84.4 | — | |
| GPipeParams=556M, Pre-trained=ImageNet ILSVRC20122021.04 | 84.4 | — | |
| Real et al.Protocol=Trained from scratch2018.11 | 83.9 | — | |
| EfficientNetV2-SParams=24M, Pre-trained=ImageNet ILSVRC20122021.04 | 83.2 | — | |
| DeiT-B-384Params=86M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=ViT+regularization2021.04 | 83.1 | — | |
| DeiT-BParams=86M, Pre-trained=ImageNet ILSVRC2012, Evaluation Protocol=ViT+regularization2021.04 | 81.8 | — | |
| ViT-L/16Params=306M, Pre-trained=ImageNet ILSVRC20122021.04 | 76.5 | — | |
| ViT-B/16Params=87M, Pre-trained=ImageNet ILSVRC20122021.04 | 74.9 | — | |
| ViT-B/32Params=88M, Pre-trained=ImageNet ILSVRC20122021.04 | 73.4 | — | |
| ViT-L/32Params=306M, Pre-trained=ImageNet ILSVRC20122021.04 | 71.2 | — | |
| SPP-net (SPP O-7)test scales=6, test views=96+2full2014.06 | — | 9.08 |