参数配置

$$参数配置$$

from major_models.LeNet import LeNet
import torch

# dict_label = {"airplane": 0, "automobile": 1, "bird": 2, "cat": 3, "deer": 4,"dog": 5,
#               "frog": 6, "horse": 7, "ship": 8, "truck": 9}
# 1.dict_label:类别对应表
dict_label = {"0": 0, "1": 1, "2": 2, "3": 3, "4": 4,"5": 5,
              "6": 6, "7": 7, "8": 8, "9": 9}  # 如果改了分类目标，这里需要修改

# 2.batchsize：批次大小
batchsize = 2

# 3.num_epoch：训练轮次，一般默认200
num_epoch = 2

# 4.crop_size:裁剪尺寸
crop_size = (32, 32) 

# 5.训练集的图片路径
train_image = r"./major_dataset_repo/split_data/train"  # r'./major_dataset_repo/major_collected_dataset/train/image'

# 6.验证集的图片路径
val_image = r'./major_dataset_repo/split_data/valid'

# 7.测试集的图片路径
test_image = r'./major_dataset_repo/split_data/test'

# 8.待转训练、验证和测试集的数据原文件
dataset_image = r'./major_dataset_repo/image'

# 9.path_test_model : 测试模型的路径
path_test_model = "./major_saved_models_repo/common/weights/best_model.pth"

# 10.path_predict_model : predict模型的路径
path_predict_model = "./major_saved_models_repo/common/weights/best_model.pth"

# 11.模型的保存路径
path_saved_model = './major_saved_models_repo/common/weights/best_model.pth'

# 12.指定设备
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')

# （norm_mean，norm_std）：数据集的均值和标准差
norm_mean = [0.45115253, 0.48260283, 0.49052352]
norm_std = [0.26216552, 0.24431673, 0.24694261]

#15.model:模型的选择
model = LeNet(num_classes=6,num_linear=44944)

深度学习的输入数据集为什么要做均值化和标准化处理

计算pytorch标准化(Normalize)所需要数据集的均值和方差

import random
from PIL import Image
from torch.utils.data import Dataset
import numpy as np
import torchvision.transforms as transforms
from torch.utils.data import DataLoader
import os

random.seed(1)
# dict_label:类别对应表
# dict_label = {"airplane": 0, "automobile": 1, "bird": 2, "cat": 3, "deer": 4,"dog": 5,
#               "frog": 6, "horse": 7, "ship": 8, "truck": 9}
dict_label = {"0": 0, "1": 1, "2": 2, "3": 3, "4": 4,"5": 5,
              "6": 6, "7": 7, "8": 8, "9": 9}  # 如果改了分类目标，这里需要修改



def get_img_info(data_dir):
    data_info = list()
    for root, dirs, _ in os.walk(data_dir):
        # 遍历类别
        for sub_dir in dirs:
            img_names = os.listdir(os.path.join(root, sub_dir))
            img_names = list(filter(lambda x: x.endswith('.png'), img_names)) # 过滤，剩下.png结尾的文件名
            # 遍历图片
            for i in range(len(img_names)):
                img_name = img_names[i]
                path_img = os.path.join(root, sub_dir, img_name) # 完整图片路径
                label = dict_label[sub_dir] # 获取当前图片的标签
                data_info.append((path_img, int(label))) # 返回 [(path_img1,label1),(path_img2,label2),...]

    return data_info

class MyDataset(Dataset):
    def __init__(self, data_dir, transform=None):
        self.label_name = dict_label
        self.data_info = get_img_info(data_dir)  # data_info存储所有图片路径和标签
        self.transform = transform

    def __getitem__(self, index):
        path_img, label = self.data_info[index]
        img = Image.open(path_img).convert('RGB')     # 0~255
        if self.transform is not None:
            img = self.transform(img)   # 在这里做transform，转为tensor等等
        return img, label

    def __len__(self):
        return len(self.data_info)


#  指定计算mean和std的图像数据集路径
train_dir = os.path.join('.', 'split_data',"train")

#  图像预处理
train_transform = transforms.Compose([
    transforms.Resize((32, 32)), # 可以改成你图片近似大小或者模型要求大小
    transforms.ToTensor(),
])

train_data = MyDataset(data_dir=train_dir, transform=train_transform)
train_loader = DataLoader(dataset=train_data, batch_size=3000, shuffle=True) # 3000张图片的mean std
train = iter(train_loader).next()[0]  # 3000张图片的mean、std
train_mean = np.mean(train.numpy(), axis=(0, 2, 3))
train_std = np.std(train.numpy(), axis=(0, 2, 3))

print("train_mean:",train_mean)
print("train_std:",train_std)

pytorch的transforms.Normalize(mean,std)是每个通道减去均值，除以标准差获得归一化图像

import torch
from torchvision import transforms

C= 2
H= 2
W= 2
arry = torch.arange(C*H*W,dtype=torch.float32).view([C,H,W])
print('输入矩阵：',arry)
 
mean = [2,2]
std  = [2,2]
# transforms.Normalize
n = transforms.Normalize(mean=mean,std=std)
print('pytorch标准化：',n(arry))
 
# 公式归一化:(arry[c]-mean[c]) / std[c] 
print('公式标准化:')
for c in range(C):
    print((arry[c]-mean[c]) / std[c])

numpy中的mean和std中axis的使用