从零搭建视觉模型 02：Python 起步，写给 C/C++ 开发者

def image_info(width, height, channels=3):
    pixels = width * height
    return {
        "width": width,
        "height": height,
        "channels": channels,
        "pixels": pixels,
        "bytes": pixels * channels,
    }


def main():
    image = image_info(3, 2)
    print(image)
    print(
        f"宽={image['width']}, 高={image['height']}, "
        f"像素数={image['pixels']}, 字节数={image['bytes']}"
    )
    print(image_info(4, 5, channels=1))


if __name__ == "__main__":
    main()

{'width': 3, 'height': 2, 'channels': 3, 'pixels': 6, 'bytes': 18}
宽=3, 高=2, 像素数=6, 字节数=18
{'width': 4, 'height': 5, 'channels': 1, 'pixels': 20, 'bytes': 20}

def clip_channels(values):
    new_values = values.copy()
    for i in range(len(values)):
        if values[i] < 0:
            new_values[i] = 0
        elif values[i] > 255:
            new_values[i] = 255
    return new_values


def main():
    values = [-20, 0, 128, 255, 300]
    result = clip_channels(values)
    print(result)
    print(values)

    result[0] = 100
    print(result)
    print(values)

    print(clip_channels([]))
    normal = [10, 20, 30]
    copied = clip_channels(normal)
    print(copied)
    print(copied is normal)


if __name__ == "__main__":
    main()

[0, 0, 128, 255, 255]
[-20, 0, 128, 255, 300]
[100, 0, 128, 255, 255]
[-20, 0, 128, 255, 300]
[]
[10, 20, 30]
False

def select_images(images, min_pixels=6):
    if images is None:
        return []

    new_images = []
    for image in images:
        if image["width"] * image["height"] >= min_pixels:
            new_images.append(image["name"])
    return new_images


def main():
    images = [
        {"name": "a.ppm", "width": 3, "height": 2},
        {"name": "b.ppm", "width": 2, "height": 2},
        {"name": "c.ppm", "width": 4, "height": 3},
    ]

    print(select_images(images))
    print(select_images(images, min_pixels=12))
    print(select_images(images, min_pixels=13))
    print(select_images([], min_pixels=6))


if __name__ == "__main__":
    main()

['a.ppm', 'c.ppm']
['c.ppm']
[]
[]

from pathlib import Path


class PPMFileInfo:
    def __init__(self, path):
        self.path = path

    def find(self):
        content = self.path.read_text(encoding="ascii")
        parts = content.split()

        self.typeinfo = parts[0]
        self.width = int(parts[1])
        self.height = int(parts[2])
        self.maxval = int(parts[3])

        # 前四项是文件头，剩余项是通道值
        values = []
        for value in parts[4:]:
            values.append(int(value))

        pixel_count = len(values) // 3
        last_pixel = values[-3:]

        print(f"格式: {self.typeinfo}")
        print(f"宽: {self.width}")
        print(f"高: {self.height}")
        print(f"通道最大值: {self.maxval}")
        print(f"像素数: {pixel_count}")
        print(f"最后一个像素: {last_pixel}")


def main():
    path = Path(__file__).resolve().parent / "sample.ppm"
    try:
        info = PPMFileInfo(path)
        info.find()
    except FileNotFoundError:
        print("图片文件不存在")


if __name__ == "__main__":
    main()

格式: P3
宽: 3
高: 2
通道最大值: 255
像素数: 6
最后一个像素: [255, 128, 0]

project_dir = Path(__file__).resolve().parent.parent
path = (
    project_dir
    / "content"
    / "posts"
    / "AI-深度学习"
    / "从零搭建视觉模型"
    / "02-Python起步"
    / "code"
    / "sample.ppm"
)

class ImageInfo:
    def __init__(self, width, height, channels=3):
        if width <= 0 or height <= 0 or channels <= 0:
            raise ValueError("尺寸和通道数必须为正数")
        self.width = width
        self.height = height
        self.channels = channels

    def pixel_count(self):
        return self.width * self.height

    def value_count(self):
        return self.pixel_count() * self.channels

    def resize(self, width, height):
        if width <= 0 or height <= 0:
            raise ValueError("新尺寸必须为正数")
        self.width = width
        self.height = height


def main():
    first = ImageInfo(3, 2)
    second = ImageInfo(2, 2, channels=1)

    print(first.pixel_count(), first.value_count())
    first.resize(5, 4)
    print(first.pixel_count(), first.value_count())
    print(second.pixel_count(), second.value_count())

    try:
        first.resize(0, 4)
    except ValueError as e:
        print(f"更新失败: {e}")
    print(f"更新失败后的像素数: {first.pixel_count()}")

    try:
        ImageInfo(-1, 2)
    except ValueError as e:
        print(f"初始化失败: {e}")


if __name__ == "__main__":
    main()

6 18
20 60
4 4
更新失败: 新尺寸必须为正数
更新失败后的像素数: 20
初始化失败: 尺寸和通道数必须为正数

from pathlib import Path

import numpy as np
from PIL import Image


def main():
    path = Path(__file__).resolve().parent / "sample.ppm"
    with Image.open(path) as source:
        image = source.convert("RGB")
    pixels = np.array(image, dtype=np.uint8, copy=True)

    print(pixels.shape)
    print(f"第二行、第三列像素: {pixels[1, 2].tolist()}")
    print("G 通道:")
    print(pixels[:, :, 1])

    original = pixels.copy()
    new_pixels = pixels.copy()
    new_pixels[:, :, 2] = 0

    print(f"原数组: {pixels[0, 2].tolist()}")
    print(f"副本: {new_pixels[0, 2].tolist()}")
    print(f"原数组未改变: {np.array_equal(pixels, original)}")
    print(f"副本 B 通道全部为零: {np.all(new_pixels[:, :, 2] == 0)}")


if __name__ == "__main__":
    main()

(2, 3, 3)
第二行、第三列像素: [255, 128, 0]
G 通道:
[[  0 255   0]
 [255   0 128]]
原数组: [0, 0, 255]
副本: [0, 0, 0]
原数组未改变: True
副本 B 通道全部为零: True