計算資料集均值方差
阿新 • • 發佈:2020-12-09
技術標籤:深度學習
參考:
https://blog.csdn.net/weixin_41765699/article/details/100118660
對於資料集獲取簡單進行修改
import os import cv2 import numpy as np def searchDirFile(rootDir, list_, path_, endswith_): for dir_or_file in os.listdir(rootDir): filePath = os.path.join(rootDir, dir_or_file) if os.path.isfile(filePath): if os.path.basename(filePath).endswith(endswith_): temp = filePath.split('/')[-2] list_.append(temp) path_.append(filePath) else: continue elif os.path.isdir(filePath): searchDirFile(filePath, list_, path_, endswith_) else: print('not file and dir '+os.path.basename(filePath)) path = "/home/mydata" namelist = [] pathlist = [] endswith = 'Texture_8Bit.png' searchDirFile(path, namelist, pathlist, endswith) R_channel = 0 G_channel = 0 B_channel = 0 for idx in range(len(pathlist)): filename = pathlist[idx] img = cv2.imread(filename) / 255.0 R_channel = R_channel + np.sum(img[:, :, 0]) G_channel = G_channel + np.sum(img[:, :, 1]) B_channel = B_channel + np.sum(img[:, :, 2]) # 這裡(512,512)是每幅圖片的大小,所有圖片尺寸都一樣 num = len(pathlist) * 2064 * 1544 R_mean = R_channel / num G_mean = G_channel / num B_mean = B_channel / num print("R_mean is %f, G_mean is %f, B_mean is %f" % (R_mean, G_mean, B_mean)) R_channel = 0 G_channel = 0 B_channel = 0 for idx in range(len(pathlist)): filename = pathlist[idx] img = cv2.imread(filename) / 255.0 R_channel = R_channel + np.sum((img[:, :, 0] - R_mean) ** 2) G_channel = G_channel + np.sum((img[:, :, 1] - G_mean) ** 2) B_channel = B_channel + np.sum((img[:, :, 2] - B_mean) ** 2) R_var = np.sqrt(R_channel / num) G_var = np.sqrt(G_channel / num) B_var = np.sqrt(B_channel / num) print("R_var is %f, G_var is %f, B_var is %f" % (R_var, G_var, B_var))
上面必須圖片大小一樣,速度很慢
如果影象大小不一樣,且要求速度較快:
https://blog.csdn.net/PanYHHH/article/details/107896526
https://zhuanlan.zhihu.com/p/275742390
通過pytorch 每次計算每張圖的均值方差,直接計入總和,然後除以資料集數目