計算數據集的均值和方差（mean，std）

coco數據集的均值和方差（三分量順序是RGB）

mean = [0.471, 0.448, 0.408]
std = [0.234, 0.239, 0.242]

 

imagenet數據集的均值和方差（三分量順序是RGB）

mean = [0.485, 0.456, 0.406]
std = [0.229, 0.224, 0.225]

 

用opencv和numpy計算自己數據集的均值和方差 

import numpy as np
import cv2
import os

# img_h, img_w = 32, 32
img_h, img_w = 32, 48   #根據自己數據集適當調整，影響不大
means, stdevs = [], []
img_list = []

imgs_path = 'D:/database/VOCdevkit/VOC2012/JPEGImages/'
imgs_path_list = os.listdir(imgs_path)

len_ = len(imgs_path_list)
i = 0
for item in imgs_path_list:
    img = cv2.imread(os.path.join(imgs_path,item))
    img = cv2.resize(img,(img_w,img_h))
    img = img[:, :, :, np.newaxis]
    img_list.append(img)
    i += 1
    print(i,'/',len_)    

imgs = np.concatenate(img_list, axis=3)
imgs = imgs.astype(np.float32) / 255.

for i in range(3):
    pixels = imgs[:, :, i, :].ravel()  # 拉成一行
    means.append(np.mean(pixels))
    stdevs.append(np.std(pixels))

# BGR --> RGB ， CV讀取的需要轉換，PIL讀取的不用轉換
means.reverse()
stdevs.reverse()

print("normMean = {}".format(means))
print("normStd = {}".format(stdevs))

參考：https://blog.csdn.net/weixin_38533896/article/details/85951903