毕设草稿保存

参数大小

MobileViT xxs参数：

MobileViT xs参数

MobileViT s参数

MobileViT+SE模块

无SE模块时

有预训练文件

无预训练文件

有预训练文件且加SE模块之后：

无预训练文件且加了SE模块

扫描二维码关注公众号，回复： 17518964 查看本文章

MobileNetv2

epoch=60 lr=2e-5

lr=1.6e-5

MobileNetv3

CosineAnnealingWarmRestarts

    scheduler = torch.optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer,
                                                                     T_0=10,  # init_epoch to change lr
                                                                     T_mult=2,  # times
                                                                     eta_min=0,  # min of lr
                                                                     last_epoch=-1,  # default=-1
                                                                     )

MobileViT 无预训练文件

test.py

import os
import json

import torch
from PIL import Image
from torchvision import transforms
import matplotlib.pyplot as plt

from MobileViT_SE import mobile_vit_small as MobileViT_SE
from MobileViT import mobile_vit_small as MobileViT
from MobileSwin import mobile_vit_small as MobileSwin
import shutil

def main():
    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

    img_size = 224
    data_transform = transforms.Compose(
        [transforms.Resize(int(img_size * 1.14)),
         transforms.CenterCrop(img_size),
         transforms.ToTensor(),
         transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])


    # read class_indict
    json_path = './class_indices.json'
    assert os.path.exists(json_path), "file: '{}' dose not exist.".format(json_path)

    with open(json_path, "r") as f:
        class_indict = json.load(f)

    # create model load model weights
    model = MobileViT(num_classes=3).to(device)
    model_weight_path = "./result_weight/MobileViT_S_best.pth"

    model.load_state_dict(torch.load(model_weight_path, map_location=device))
    model.eval()
    class_name='Seborrheic_keratosis'
    # 设置输入和输出文件夹路径
    input_folder = "D:\Deep-learning\deep-learning-for-image-processing-master\data_set\skin_data\Test"+'\\'+class_name  # 输入文件夹，包含待处理的图片
    output_folder = "D:\Deep-learning\deep-learning-for-image-processing-master\data_set\Test"+'\\'+class_name  # 输出文件夹，用于存储符合条件的图片

    # 确保输出文件夹存在
    if not os.path.exists(output_folder):
        os.makedirs(output_folder)

    # 遍历输入文件夹中的所有文件
    for filename in os.listdir(input_folder):
        # 检查图片是否是该类别"
        img_path = os.path.join(input_folder, filename)
        img = Image.open(img_path)
        # plt.imshow(img)
        img = data_transform(img)
        # expand batch dimension
        img = torch.unsqueeze(img, dim=0)

        with torch.no_grad():
            # predict class
            output = torch.squeeze(model(img.to(device))).cpu()
            predict = torch.softmax(output, dim=0)
            predict_cla = torch.argmax(predict).numpy()

        print_res = "class: {}   prob: {:.3}".format(class_indict[str(predict_cla)],
                                                     predict[predict_cla].numpy())
        id=''
        npmax=0
        plt.title(print_res)
        for i in range(len(predict)):
            print("class: {:10}   prob: {:.3}".format(class_indict[str(i)],
                                                      predict[i].numpy()))
            if npmax<predict[i].numpy():
                npmax=predict[i].numpy()
                id=class_indict[str(i)]

        print(id)
        if id == class_name:
            # 构建输入和输出文件的完整路径
            input_filepath = os.path.join(input_folder, filename)
            output_filepath = os.path.join(output_folder, filename)

            # 将符合条件的文件复制到输出文件夹
            shutil.copyfile(input_filepath, output_filepath)

            print(f"Copied file: {filename}")


if __name__ == '__main__':
    main()