pycvt 0.0.1.post1__py3-none-any.whl

pycvt/__init__.py

@@ -0,0 +1,3 @@
+from importlib.resources import files
+
+red_dog_base_path = files("reddog")

pycvt/clolors/colors.py

@@ -0,0 +1,84 @@
+import colorsys
+import hashlib
+from functools import lru_cache
+
+_COLORS = {
+    # 绿色
+    'green': (0, 255, 0),
+    # 红色
+    'red': (0, 0, 255),
+    # 蓝色
+    'blue': (255, 0, 0),
+    # 黄色
+    'yellow': (0, 255, 255),
+    # 紫色
+    'purple': (255, 0, 255),
+    # 青色
+    'cyan': (255, 255, 0),
+    # 黑色
+    'black': (0, 0, 0),
+    # 白色
+    'white': (255, 255, 255),
+    # 灰色
+    'gray': (128, 128, 128),
+    # 深灰色
+    "GhostWhite": (248, 248, 255)
+}
+
+
+def get_luminance(bgr: tuple[int, int, int]) -> float:
+    # BGR to RGB
+    b, g, r = bgr
+    return 0.299 * r + 0.587 * g + 0.114 * b  # ITU-R BT.601
+
+
+def ensure_contrast(text_color: tuple[int, int, int], background_color: tuple[int, int, int]) -> tuple[int, int, int]:
+    text_lum = get_luminance(text_color)
+    bg_lum = get_luminance(background_color)
+
+    contrast = abs(text_lum - bg_lum)
+    if contrast < 128:  # 亮度差太小，自动调深背景
+        if text_lum > 128:
+            return (32, 32, 32)  # 改成深灰色背景
+        else:
+            return (255, 255, 255)  # 改成白色背景
+    return background_color
+
+
+def get_vibrant_color_from_key(key: str) -> tuple[int, int, int]:
+    # 使用 MD5 哈希获取稳定值
+    md5_bytes = hashlib.md5(key.encode()).digest()  # 16 字节
+    hash_bytes = md5_bytes[-3:]  # 从中随机选 3 个字节
+
+    # 使用前 3 字节作为 HSV 分量来源
+    h = hash_bytes[0] / 255.0  # Hue: [0, 1]
+    s = 0.8 + (hash_bytes[1] % 50) / 100.0  # Saturation: [0.8, 1.3] (上限1.0后裁剪)
+    v = 0.8 + (hash_bytes[2] % 50) / 100.0  # Value: [0.8, 1.3] (上限1.0后裁剪)
+
+    s = min(s, 1.0)
+    v = min(v, 1.0)
+
+    # 转为 RGB，再转为 BGR（OpenCV格式）
+    r, g, b = colorsys.hsv_to_rgb(h, s, v)
+    return int(b * 255), int(g * 255), int(r * 255)  # OpenCV 使用 BGR
+
+
+@lru_cache(maxsize=1000)
+def getcolor(key="red", bgr=True):
+    """
+    Get a color by key, either as BGR tuple or hex string.
+
+    这是一个稳定的方法，相同的 key 总是返回相同的颜色。
+    :param key:
+    :param bgr:
+    :return:
+    """
+    key = str(key)
+
+    if key in _COLORS:
+        cl_bgr = _COLORS[key]
+    else:
+        b, g, r = get_vibrant_color_from_key(key)
+        cl_bgr = (b, g, r)
+        _COLORS[key] = cl_bgr
+    return cl_bgr if bgr else f'#{r:02x}{g:02x}{b:02x}'

pycvt/paster/paste_image.py

@@ -0,0 +1,54 @@
+from copy import deepcopy
+
+import numpy as np
+
+from pycvt.utils.image_utils import convert_rgba, blend_rgba
+
+
+def paste_image(
+        bg_img,
+        fg_img,
+        position=None,
+        alpha=1.0,
+):
+    """
+    fg_img will be pasted onto bg_img at the specified position with the given alpha blending factor.
+
+    1. If position is None, a random position will be chosen within the bounds of the background image.
+    2. IF position + fg_img exceeds the bounds of bg_img, it will be clipped to fit within bg_img.
+
+    :param bg_img: Background image, can be a numpy array or a PIL image.
+    :param fg_img: Foreground image, can be a numpy array or a PIL image.
+    :param position: Tuple (y, x) indicating where to paste the foreground image on the background image.
+    :param alpha: Blending factor for the foreground image, where 0.0 is fully transparent and 1.0 is fully opaque.
+    :return: Tuple containing the modified background image and the bounding box of the pasted foreground image.
+    """
+
+    bg_img = deepcopy(bg_img)
+    fg_img = deepcopy(fg_img)
+
+    bg_img = convert_rgba(bg_img)
+    fg_img = convert_rgba(fg_img)
+
+    img_h, img_w, = bg_img.shape[:2]
+    fg_h, fg_w, = fg_img.shape[:2]
+
+    alpha = np.clip(alpha, 0.0, 1.0)
+
+    if position is None:
+        position = np.random.randint(0, img_h - fg_h // 2), np.random.randint(0, img_w - fg_w // 2),
+
+    ymin, xmin = position
+    ymax, xmax = min(img_h, position[0] + fg_h), min(img_w, position[1] + fg_w)
+
+
+    paste_size= (xmax - xmin, ymax - ymin)
+
+    bg_crop = bg_img[ymin:ymax, xmin:xmax]
+    fg_crop = fg_img[:paste_size[1], :paste_size[0]]
+
+    blended = blend_rgba(bg_crop, fg_crop, fgweight=alpha)
+
+    bg_img[ymin:ymax, xmin:xmax] = blended
+
+    return bg_img, (xmin, ymin, xmax, ymax)

pycvt/utils/__init__.py

@@ -0,0 +1 @@
+from . import image_utils

pycvt/utils/image_utils.py

@@ -0,0 +1,152 @@
+from copy import deepcopy
+
+import cv2
+import numpy as np
+
+
+def convert_rgba(img: np.ndarray) -> np.ndarray:
+    """
+    支持灰度图、RGB图和RGBA图，统一输出 RGBA。
+    已经是 RGBA 的直接返回。
+
+    参数：
+        img (np.ndarray): 输入图像
+
+    返回：
+        np.ndarray: RGBA 图像
+    """
+    if len(img.shape) == 2:
+        # 灰度图 (H, W)
+        img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGBA)
+    elif img.shape[2] == 3:
+        # RGB (H, W, 3)
+        img = cv2.cvtColor(img, cv2.COLOR_RGB2RGBA)
+    elif img.shape[2] == 4:
+        # 已经是 RGBA，直接返回
+        return img
+    else:
+        raise ValueError(f"只支持灰度图、RGB和RGBA图，当前通道数: {img.shape[2] if len(img.shape) > 2 else 1}")
+
+    return img
+
+
+def crop_image(img: np.ndarray, x: int, y: int, crop_w: int, crop_h: int) -> np.ndarray:
+    """
+    从图像中裁剪指定区域，自动限制不会超出图像边界。
+
+    参数：
+        img (np.ndarray): 输入图像，shape (H, W, C) 或 (H, W)
+        x (int): 裁剪区域左上角的横坐标（列）
+        y (int): 裁剪区域左上角的纵坐标（行）
+        crop_w (int): 裁剪宽度
+        crop_h (int): 裁剪高度
+
+    返回：
+        np.ndarray: 裁剪后的图像区域
+    """
+    img_h, img_w = img.shape[:2]
+
+    # 限制裁剪坐标和尺寸不超过图像范围
+    x_end = min(img_w, x + crop_w)
+    y_end = min(img_h, y + crop_h)
+
+    # 防止坐标越界（x, y不能小于0）
+    x = max(0, x)
+    y = max(0, y)
+
+    return img[y:y_end, x:x_end]
+
+
+def blend_rgba(
+        bg_crop: np.ndarray,
+        fg_crop: np.ndarray,
+        **kwargs,
+) -> np.ndarray:
+    """
+    只替换 fg_crop 中 alpha > 0 的区域，直接覆盖 bg_crop 对应像素，
+    其他部分保留 bg_crop。
+
+    要求 bg_crop 和 fg_crop 形状相同，且均为 RGBA。
+    """
+    if bg_crop.shape != fg_crop.shape or bg_crop.shape[2] != 4:
+        raise ValueError("输入图像需为形状相同的 RGBA 图像")
+
+    out = bg_crop.copy()
+
+    alpha_mask = fg_crop[:, :, 3] > 0  # bool mask，fg不透明区域
+
+    for c in range(4):
+        channel_bg = out[:, :, c]
+        channel_fg = fg_crop[:, :, c]
+        channel_bg[alpha_mask] = channel_fg[alpha_mask]
+
+    return out
+
+
+def get_opaque_bounding_box(rgba_img: np.ndarray, alpha_threshold: int = 0):
+    """
+    给定 RGBA 图，返回不透明区域的外接矩形 (top, left, bottom, right)。
+
+    alpha_threshold：不透明的阈值，默认为0，表示alpha>0的像素算不透明。
+
+    如果全透明，则返回 None。
+    """
+    if rgba_img.shape[2] != 4:
+        raise ValueError("输入图像必须是 RGBA 格式")
+
+    alpha = rgba_img[:, :, 3]
+    mask = alpha > alpha_threshold
+
+    if not np.any(mask):
+        return None  # 全透明无外接矩形
+
+    coords = np.argwhere(mask)  # 找出所有不透明像素的坐标，格式为 (y, x)
+    ymin, xmin = coords.min(axis=0)
+    ymax, xmax = coords.max(axis=0)
+
+    return (xmin, ymin, xmax, ymax)
+
+
+def rotate_image_with_bound(image: np.ndarray, angle: float = None) -> np.ndarray:
+    """
+    旋转图像，自动扩充尺寸以适应整个旋转后的图像。
+    RGB 背景填白，RGBA 背景填透明。
+
+    参数:
+        image: 输入图像 (H, W, 3) 或 (H, W, 4)
+        angle: 逆时针旋转角度，单位度，如果为 None，则随机选择一个角度。
+
+    返回:
+        旋转后自动扩充尺寸的图像
+    """
+
+    image = deepcopy(image)
+
+    if angle is None:
+        angle = np.random.randint(-180, 180)
+
+    (h, w) = image.shape[:2]
+    (cx, cy) = (w // 2, h // 2)
+
+    M = cv2.getRotationMatrix2D((cx, cy), -angle, 1.0)
+
+    cos = np.abs(M[0, 0])
+    sin = np.abs(M[0, 1])
+
+    new_w = int((h * sin) + (w * cos))
+    new_h = int((h * cos) + (w * sin))
+
+    M[0, 2] += (new_w / 2) - cx
+    M[1, 2] += (new_h / 2) - cy
+
+    # 判断通道数，设置填充颜色
+    if image.shape[2] == 3:
+        border_val = (255, 255, 255)  # 白色填充
+    elif image.shape[2] == 4:
+        border_val = (0, 0, 0, 0)  # 透明填充
+    else:
+        raise ValueError("只支持 RGB 或 RGBA 图像")
+
+    rotated = cv2.warpAffine(image, M, (new_w, new_h), flags=cv2.INTER_LINEAR,
+                             borderMode=cv2.BORDER_CONSTANT, borderValue=border_val)
+    return rotated

pycvt/vision/plot_boxes.py

@@ -0,0 +1,100 @@
+from copy import deepcopy
+
+import cv2
+import numpy
+import numpy as np
+from PIL import Image, ImageDraw, ImageFont
+from easyfont import getfont
+
+from pycvt.clolors.colors import getcolor, ensure_contrast
+
+
+def draw_text(
+        img,
+        text,
+        position,
+        font_path=None,
+        font_size=None,
+        text_color=None,
+        background_color=None
+):
+    """
+    在图像上绘制文本
+    :param img:  图像对象（OpenCV格式）
+    :param text:  要绘制的文本(受字体支持的字符)
+    :param position:  文本位置，建议使用 xmax, ymax 作为位置 ，防止覆盖包围框
+    :param font_path: 字体路径
+    :param font_size:  字体大小
+    :param text_color:  文本颜色
+    :param background_color:  文本背景颜色
+    :return:
+    """
+    # 将 OpenCV 图像转换为 PIL 图像
+
+    if font_path is None:  # 若未指定字体路径，则使用默认字体
+        font_path = getfont()
+
+    text_color = text_color or getcolor(text)
+    background_color = background_color or getcolor("GhostWhite")
+    background_color = ensure_contrast(text_color, background_color)  # 确保文本颜色和背景颜色有足够对比度
+
+    img_pil = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
+    font = ImageFont.truetype(font_path, font_size)
+    draw = ImageDraw.Draw(img_pil)
+
+    spacing = font_size // 3
+    half_spacing = spacing // 2
+
+    text_bbox = draw.textbbox((0, 0), text, font=font, font_size=font_size, spacing=spacing)
+    text_bbox_h = text_bbox[3]
+    text_bbox_w = text_bbox[2]
+
+    bgpos = (
+        position[0] - text_bbox_w - half_spacing,
+        position[1] + spacing,
+        position[0] + spacing,
+        position[1] + text_bbox_h + spacing
+    )
+
+    text_pos = position[0] - text_bbox_w, position[1] + half_spacing
+    draw.rectangle(bgpos, fill=background_color)
+    draw.text(text_pos, text, font=font, fill=text_color, spacing=spacing)
+    return cv2.cvtColor(np.array(img_pil), cv2.COLOR_RGB2BGR)
+
+
+def draw_bounding_boxes(
+        image: numpy.ndarray,
+        boxes: list,
+        labels=None,
+        colors=None,
+        width=None,
+        font=None,
+        font_size=None
+):
+    img_copy = deepcopy(image)
+
+    count = len(boxes)
+    w, h = img_copy.shape[1], img_copy.shape[0]
+
+    if colors is None:
+        colors = list(map(getcolor, labels)) if labels else [getcolor()] * count
+
+    line_width = width if width else max(int(0.0035 * min(w, h)), 2)
+    font_size = font_size if font_size else max(int(0.018 * min(w, h)), 2)
+
+    font = font if font else getfont()
+    for idx, box in enumerate(boxes):
+        color = colors[idx]
+        xmin, ymin, xmax, ymax = box
+        img_copy = cv2.rectangle(img_copy, (xmin, ymin), (xmax, ymax), color, line_width)
+        if labels:
+            label = labels[idx]
+            img_copy = draw_text(
+                img_copy,
+                label,
+                (xmax, ymax),
+                text_color=color[::-1],  # OpenCV uses BGR, PIL uses RGB
+                font_path=font,
+                font_size=font_size
+            )
+    return img_copy

pycvt/vision/utils.py

@@ -0,0 +1,25 @@
+import numpy as np
+
+
+def iou(gtboxes, dtboxes):
+    '''numpy version of calculating IoU between two set of 2D bboxes.
+
+    Args:
+        gtboxes (np.ndarray): Shape (B,4) of ..,  4 present [x1,y1,x2,y2]
+        dtboxes,np.ndarray,shape:(N,4), 4 present [x1,y1,x2,y2].
+
+    Returns:
+        np.ndarray: Shape (B,N)  .
+    '''
+
+
+    gtboxes = gtboxes[:, np.
+                      newaxis, :]  #converse gtboxes:(B,4) to gtboxes:(B,1,4)
+    ixmin = np.maximum(gtboxes[:, :, 0], dtboxes[:, 0])
+    iymin = np.maximum(gtboxes[:, :, 1], dtboxes[:, 1])
+    ixmax = np.minimum(gtboxes[:, :, 2], dtboxes[:, 2])
+    iymax = np.minimum(gtboxes[:, :, 3], dtboxes[:, 3])
+    intersection = (ixmax - ixmin + 1) * (iymax - iymin + 1)
+    union = (gtboxes[:,:,2]-gtboxes[:,:,0]+1)*(gtboxes[:,:,3]-gtboxes[:,:,1]+1)\
+            +(dtboxes[:,2]-dtboxes[:,0]+1)*(dtboxes[:,3]-dtboxes[:,1]+1)-intersection
+    return intersection / union

pycvt-0.0.1.post1.dist-info/METADATA

@@ -0,0 +1,39 @@
+Metadata-Version: 2.4
+Name: pycvt
+Version: 0.0.1.post1
+Summary: A lightweight Python toolkit for seamless image composition, color processing, and visualizing detection results.
+Author-email: ferretsprite <ferret-sprite-goon@duck.com>
+Project-URL: Homepage, https://github.com/xx025/pycvt
+Project-URL: Repository, https://github.com/xx025/pycvt
+Project-URL: Documentation, https://github.com/xx025/pycvt#readme
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: easyfont>=0.0.1.post10
+Requires-Dist: imageio>=2.37.0
+Requires-Dist: pillow>=11.2.1
+Requires-Dist: torch>=2.7.1
+Requires-Dist: torchvision>=0.22.1
+Requires-Dist: ultralytics>=8.3.151
+
+# pycvt
+
+
+
+## Install
+
+```
+pip install pycvt --upgrade
+```
+
+## dev
+
+``` 
+uv sync
+```
+
+
+## API List
+
+- [colors.py](pycvt/clolors/colors.py)
+- [paste_image](pycvt/paster/paste_image.py)
+- wait for more...

pycvt-0.0.1.post1.dist-info/RECORD

@@ -0,0 +1,15 @@
+pycvt/__init__.py,sha256=6cwlC2WpWOqng9qk-CwqHc7QQQXdS9eBJb33Kd8NX-Q,75
+pycvt/clolors/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+pycvt/clolors/colors.py,sha256=5Xd7tT_Rqd8TLuCOPGOK_XpSbybwka95Y8c-7ORPSfU,2353
+pycvt/paster/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+pycvt/paster/paste_image.py,sha256=MGhgiH1pVjHAPbFnXyWfRcjaDDY7eEr_JdwYTsvmE1A,1776
+pycvt/static/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+pycvt/utils/__init__.py,sha256=VxWtYveyd5NMI2gULOkqmZxMpOqY_ALK-nHtcsHkWJ4,25
+pycvt/utils/image_utils.py,sha256=VZFDPmZaISegmUyM1hGdcmAZytmYfG2SglLnhEEa2W0,4353
+pycvt/vision/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+pycvt/vision/plot_boxes.py,sha256=ywhLIqwRk7zDMLtnDlHpp1DkKIxqAwRl_mt32iYNVG8,3103
+pycvt/vision/utils.py,sha256=KJeD8d7vqTpfEht8YGdrGE6--tYwD-S86_kM5KqcCzg,917
+pycvt-0.0.1.post1.dist-info/METADATA,sha256=5VomFw4_TblIT3AgYA6ZNsx3kVNfwwZYgE_yJWKQqq0,877
+pycvt-0.0.1.post1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+pycvt-0.0.1.post1.dist-info/top_level.txt,sha256=GJFVu_yXlM6d0jjxa0u1kQNb-E2YRgpRDWgk2HRKI4s,6
+pycvt-0.0.1.post1.dist-info/RECORD,,

pycvt-0.0.1.post1.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (80.9.0)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

pycvt-0.0.1.post1.dist-info/top_level.txt

@@ -0,0 +1 @@
+pycvt