目标检测：计算自建数据集的平均值和方差

方式一：

import numpy as np
import cv2
import os


paths = '这里输入待计算图片的上层路径'

means = [0, 0, 0]
stdevs = [0, 0, 0]

index = 1
num_imgs = 0

for file in os.listdir(paths):
 if (file.endswith('.jpg')):
        print('{}/{}'.format(index, len(file)))
        index += 1
        a = os.path.join(paths,file)
        # print(a[:-1])
        num_imgs += 1
        img = cv2.imread(a)
        print(a)
        img = np.asarray(img)

        img = img.astype(np.float32) / 255.
        for i in range(3):
            means[i] += img[:, :, i].mean()
            stdevs[i] += img[:, :, i].std()
print(num_imgs)
means.reverse()
stdevs.reverse()

means = np.asarray(means) / num_imgs
stdevs = np.asarray(stdevs) / num_imgs

print("normMean = {}".format(means))
print("normStd = {}".format(stdevs))
print('transforms.Normalize(normMean = {}, normStd = {})'.format(means, stdevs))

方式二

ImageFolder
一个通用的数据加载器，数据集中的数据以以下方式组织

root/dog/xxx.png
root/dog/xxy.png
root/dog/xxz.png

root/cat/123.png
root/cat/nsdf3.png
root/cat/asd932_.png

dset.ImageFolder(root=“root folder path”, [transform, target_transform])
他有以下成员变量:

self.classes - 用一个list保存 类名
self.class_to_idx - 类名对应的 索引
self.imgs - 保存(img-path, class) tuple的list

import numpy as np
from PIL import Image
import torchvision
from time import time
from tqdm import tqdm

def compute_mean_and_std(dataset):
    # 输入PyTorch的dataset，输出均值和标准差
    mean_r = 0
    mean_g = 0
    mean_b = 0
    print("计算均值>>>")
    for img_path, _ in tqdm(dataset,ncols=80):
      img=Image.open(img_path)
      img = np.asarray(img) # change PIL Image to numpy array
      print(img.shape)
      if(len(img.shape)==2):
          image = np.expand_dims(img, axis=2)
          img = np.concatenate((image, image, image), axis=-1)
      mean_b += np.mean(img[:, :, 0])
      mean_g += np.mean(img[:, :, 1])
      mean_r += np.mean(img[:, :, 2])

    mean_b /= len(dataset)
    mean_g /= len(dataset)
    mean_r /= len(dataset)

    diff_r = 0
    diff_g = 0
    diff_b = 0

    N = 0
    print("计算方差>>>")
    for img_path, _ in tqdm(dataset,ncols=80):
      img=Image.open(img_path)
      img = np.asarray(img)
      if (len(img.shape) == 2):
          image = np.expand_dims(img, axis=2)
          img = np.concatenate((image, image, image), axis=-1)
      diff_b += np.sum(np.power(img[:, :, 0] - mean_b, 2))
      diff_g += np.sum(np.power(img[:, :, 1] - mean_g, 2))
      diff_r += np.sum(np.power(img[:, :, 2] - mean_r, 2))

      N += np.prod(img[:, :, 0].shape)

    std_b = np.sqrt(diff_b / N)
    std_g = np.sqrt(diff_g / N)
    std_r = np.sqrt(diff_r / N)

    mean = (mean_b.item() / 255.0, mean_g.item() / 255.0, mean_r.item() / 255.0)
    std = (std_b.item() / 255.0, std_g.item() / 255.0, std_r.item() / 255.0)
    return mean, std

path = "输入保存图片的上层路径"
val_data = torchvision.datasets.ImageFolder(root=path)

time_start =time()
val_mean,val_std=compute_mean_and_std(val_data.imgs)
time_end=time()
print("计算消耗时间：", round(time_end - time_start, 4), "s")
print("数据集的平均值：{}".format(val_mean))
print("数据集的方差：{}".format(val_std))

这里贴一下用torchvision.datasets.ImageFolder一下可能遇到的错误的解决方案

torchvision.datasets.ImageFolder把文件夹自动的转化为0，1，2…等类别号，按顺序访问。