文件 | 说明 |
---|---|
train.py | 训练 |
predict.py | 预测 |
test.py | 用于测试,实际文件没有用 |
config.py | 配置文件,设置配置参数 |
utils/utils.py | 其他代码 |
log/*.txt | 保存实验结果 |
pre_pth | 预训练模型 |
checkpoint | 训练好的模型 |
Method | Pretrained | val_acc | test_acc | precision | recall | macro-F1 |
---|---|---|---|---|---|---|
AlexNet | √ | 0.669 | 0.641 | 0.675 | 0.632 | 0.642 |
AlexNet | 0.584 | 0.526 | 0.564 | 0.519 | 0.529 | |
VGG11 | √ | 0.723 | 0.687 | 0.693 | 0.684 | 0.68 |
VGG11 | 0.602 | 0.576 | 0.586 | 0.569 | 0.57 | |
VGG13 | √ | 0.729 | 0.693 | 0.74 | 0.68 | 0.691 |
VGG13 | 0.601 | 0.544 | 0.568 | 0.532 | 0.529 | |
VGG16 | √ | 0.734 | 0.707 | 0.748 | 0.692 | 0.702 |
VGG16 | 0.596 | 0.559 | 0.585 | 0.55 | 0.552 | |
VGG19 | √ | 0.731 | 0.708 | 0.722 | 0.7 | 0.702 |
VGG19 | 0.587 | 0.547 | 0.567 | 0.538 | 0.541 | |
ResNet18 | √ | 0.714 | 0.686 | 0.691 | 0.683 | 0.687 |
ResNet18 | 0.609 | 0.576 | 0.598 | 0.568 | 0.576 | |
ResNet34 | √ | 0.75 | 0.729 | 0.74 | 0.719 | 0.72 |
ResNet34 | 0.617 | 0.571 | 0.59 | 0.563 | 0.567 | |
ResNet50 | √ | 0.733 | 0.707 | 0.715 | 0.698 | 0.701 |
ResNet50 | 0.588 | 0.567 | 0.583 | 0.561 | 0.567 | |
ResNet101 | √ | 0.723 | 0.708 | 0.719 | 0.706 | 0.71 |
ResNet101 | 0.584 | 0.553 | 0.581 | 0.538 | 0.542 | |
MobileNetV2 | √ | 0.729 | 0.722 | 0.733 | 0.717 | 0.72 |
MobileNetV2 | 0.534 | 0.495 | 0.518 | 0.491 | 0.5 | |
MobileNetV3 | √ | 0.711 | 0.667 | 0.68 | 0.658 | 0.664 |
MobileNetV3 | 0.504 | 0.474 | 0.482 | 0.464 | 0.468 | |
MobileNetV3-large | √ | 0.735 | 0.720 | 0.734 | 0.713 | 0.718 |
-
Code
-
Clone this repository in the directory (
Root/classification
):git clone https://github.com/iOPEN-Yang/classification.git
-
-
Pretrained and Fine-tuned pth-file downloading (Scan through WeChat)