EfficientNet-B0 was introduced by Tan et al. in 2019. Starting from the baseline EfficientNet-B0, they applied compound scaling method to scale it up to obtain EfficientNet-B1 to B7, which achieve much better accuracy and efficiency than previous ConvNets.
| Model | Download | #Params (M) | #FLOPs (B) | Top-1 accuracy (%) | Top-5 accuracy (%) |
|---|---|---|---|---|---|
| EfficientNet-B0 | 22MB | 5.3 | 0.39 | 77.1 | 93.3 |
- EfficientNet-B0 is from the paper titled EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.
- This pre-train onnx model is converted from a pytorch model, which is from the lukemelas/EfficientNet-PyTorch.
Apache 2.0