efficientnet_pytorch

实现风格与torchvision.models中其他网络的类似
(The implementation style is similar to other networks in torchvision.models)

Reference

1. 论文(paper):
   https://arxiv.org/pdf/1905.11946.pdf

2. 代码参考(reference code):
   https://github.com/lukemelas/EfficientNet-PyTorch

3. 使用的预训练模型来源(Source of the pre-training model used):
   https://github.com/lukemelas/EfficientNet-PyTorch
   由于我对模型进行了调整，所以我修改了原预训练模型的state_dict的key值，并进行了发布.
   Since I adjusted the model, I modified the state_dict key value of the original pretraining model and release it.

权重见 release. 或在百度云中下载:
链接：https://pan.baidu.com/s/1HzAZZoJNIoii7uDsnPExHQ
提取码：4sb5

4. labels_map.txt来源(Source of labels_map.txt):
   https://github.com/lukemelas/EfficientNet-PyTorch/blob/master/examples/simple/labels_map.txt

使用方式(How to use)

1. 预测图片(Predict images)

python3 std_pred_image.py

2. 简单的训练例子(Simple training examples)

python3 train_example.py

3. train

python3 make_dataset.py
python3 train.py
python3 pred_image.py

性能

见docs/文件夹

运行环境(environment)

torch 1.8.1
torchvision 0.9.0