时间戳

首页 > 解决方案 > 将 numba 添加到 python 脚本

python - 将 numba 添加到 python 脚本

问题描述

我试图将 numba 应用于以下脚本以减少运行时间，但没有成功。关于如何做到这一点的任何建议？我在 spyder/anaconda 中使用 python 3.7 工作。

import timeit
start = timeit.default_timer()


from PIL import Image, ImageDraw # Importerer biblioteket


# Load image:
input_image = Image.open("beatles.jpg") # Loads an image from the hard drive
input_pixels = input_image.load()

# Box Blur kernel
box_kernel = [[1 / 9, 1 / 9, 1 / 9], [1 / 9, 1 / 9, 1 / 9], [1 / 9, 1 / 9, 1 / 9]]

# Select kernel here:
kernel = box_kernel

# Middle of the kernel
offset = len(kernel) // 2

# Creating the output image
output_image = Image.new("RGB", input_image.size)
draw = ImageDraw.Draw(output_image)


# Beregne konvolusjon mellom intensity (original bilde) og  kernel 
for x in range(offset, input_image.width - offset):
    for y in range(offset, input_image.height - offset):
        acc = [0, 0, 0]
        for a in range(len(kernel)):
            for b in range(len(kernel)):
                xn = x + a - offset
                yn = y + b - offset
                pixel = input_pixels[xn, yn]
                acc[0] += pixel[0] * kernel[a][b]
                acc[1] += pixel[1] * kernel[a][b]
                acc[2] += pixel[2] * kernel[a][b]

draw.point((x, y), (int(acc[0]), int(acc[1]), int(acc[2])))

output_image.save("Beatles box blur.png")

stop = timeit.default_timer()

print('Time: ', stop - start)

标签： pythonanaconda

解决方案

我在这里做了一些快速的事情，希望它可以让你开始：

import time
import numpy as np

from numba import jit


def numba_off(offset, width, height, kernel):
    for x in range(offset, width - offset):
        for y in range(offset, height - offset):
            acc = [0, 0, 0]
            for a in range(len(kernel)):
                for b in range(len(kernel)):
                    pixel = [1, 1, 1]
                    acc[0] += pixel[0] * kernel[a][b]
                    acc[1] += pixel[1] * kernel[a][b]
                    acc[2] += pixel[2] * kernel[a][b]


@jit(nopython=True)
def numba_on(offset, width, height, kernel):
    for x in range(offset, width - offset):
        for y in range(offset, height - offset):
            acc = [0, 0, 0]
            for a in range(len(kernel)):
                for b in range(len(kernel)):
                    pixel = [1, 1, 1]
                    acc[0] += pixel[0] * kernel[a][b]
                    acc[1] += pixel[1] * kernel[a][b]
                    acc[2] += pixel[2] * kernel[a][b]


box_kernel = np.array([[1 / 9, 1 / 9, 1 / 9], [1 / 9, 1 / 9, 1 / 9], [1 / 9, 1 / 9, 1 / 9]])
kernel = box_kernel
offset = len(kernel) // 2

start = time.time()
numba_off(offset, 1000, 1000, kernel)
end = time.time()
print("Elapsed (without) = %s" % (end - start))

start = time.time()
numba_on(offset, 1000, 1000, kernel)
end = time.time()
print("Elapsed (with compilation) = %s" % (end - start))
start = time.time()
numba_on(offset, 1000, 1000, kernel)
end = time.time()
print("Elapsed (after compilation) = %s" % (end - start))

您应该能够在此基础上再接再厉。阅读这样的内容可能对您有好处：5 分钟 numba 指南

推荐阅读