PIL和cv2读图片速度对比

使用cornell grasp dataset测试，我之前做实验删过一些，剩下878张图像。

程序

# -*- coding: UTF-8 -*-
"""===============================================
@Author : wangdx
@Date   : 2020/8/5 15:26
==============================================="""

from PIL import Image
import cv2
import numpy as np
import time
import os

path = 'E:/research/dataset/grasp/cornell/wdx/TriangleLabel/label'

file_dirs = []
for file in os.listdir(path):
    file_dirs.append(os.path.join(path, file))

# PIL测试
start_time = time.time()
for file in file_dirs:
    im = np.array(Image.open(file))
pil_time = time.time() - start_time

# cv
start_time = time.time()
for file in file_dirs:
    im = cv2.imread(file)
cv_time = time.time() - start_time

print('image num: ', len(file_dirs))
print('PIL: sum time: {:.5f}'.format(pil_time))
print('cv: sum time: {:.5f}'.format(cv_time))

结果

image num:  878
PIL: sum time: 11.54827
cv: sum time: 4.84759