Set line width and label font size proportional to image size

README.md

@@ -1,3 +1,7 @@
+
+Fork from https://github.com/nalepae/bounding-box/tree/master. I've set the line width and label size proportional to the image size. Helpfull for large images. I've opened a pull request to the original repo, but it appears to be non-active. Hence I'll use my own fork. Original documentation below.
+
+
 # Bounding Box
 **Bounding Box** is a library to plot pretty bounding boxes with a simple Python API.
 

bounding_box/bounding_box.py

@@ -8,6 +8,7 @@
 from PIL import ImageFont
 import numpy as _np
 from hashlib import md5 as _md5
+import string as _string
 
 _LOC = _path.realpath(_path.join(_os.getcwd(),_path.dirname(__file__)))
 
@@ -36,17 +37,35 @@
 _DEFAULT_COLOR_NAME = "green"
 
 _FONT_PATH = _os.path.join(_LOC, "Ubuntu-B.ttf")
-_FONT_HEIGHT = 15
-_FONT = ImageFont.truetype(_FONT_PATH, _FONT_HEIGHT)
+
+def _hex_to_rgb(color):
+    if type(color) is not str:
+        return None
+    color = color.strip()
+    if color.startswith("#"):
+        color = color[1:]
+    if len(color) == 3:
+        color = "".join(ch * 2 for ch in color)
+    if len(color) != 6 or any(ch not in _string.hexdigits for ch in color):
+        return None
+    try:
+        return tuple(int(color[i:i + 2], 16) for i in (0, 2, 4))
+    except ValueError:
+        return None
+
+def _get_contrasting_rgb(rgb):
+    r, g, b = rgb
+    luminance = 0.299 * r + 0.587 * g + 0.114 * b
+    return (255, 255, 255) if luminance < 186 else (0, 0, 0)
 
 def _rgb_to_bgr(color):
     return list(reversed(color))
 
 def _color_image(image, font_color, background_color):
     return background_color + (font_color - background_color) * image / 255
 
-def _get_label_image(text, font_color_tuple_bgr, background_color_tuple_bgr):
-    text_image = _FONT.getmask(text)
+def _get_label_image(text, font_color_tuple_bgr, background_color_tuple_bgr, font):
+    text_image = font.getmask(text)
     shape = list(reversed(text_image.size))
     bw_image = np.array(text_image).reshape(shape)
 
@@ -58,7 +77,7 @@ def _get_label_image(text, font_color_tuple_bgr, background_color_tuple_bgr):
 
     return np.concatenate(image).transpose(1, 2, 0)
 
-def add(image, left, top, right, bottom, label=None, color=None):
+def add(image, left, top, right, bottom, label=None, color=None, size=2):
     if type(image) is not _np.ndarray:
         raise TypeError("'image' parameter must be a numpy.ndarray")
     try:
@@ -79,20 +98,53 @@ def add(image, left, top, right, bottom, label=None, color=None):
 
     if type(color) is not str:
         raise TypeError("'color' must be a str")
-
-    if color not in _COLOR_NAME_TO_RGB:
-        msg = "'color' must be one of " + ", ".join(_COLOR_NAME_TO_RGB)
-        raise ValueError(msg)
-
-    colors = [_rgb_to_bgr(item) for item in _COLOR_NAME_TO_RGB[color]]
+    color = color.strip()
+
+    if color in _COLOR_NAME_TO_RGB:
+        box_rgb, text_rgb = _COLOR_NAME_TO_RGB[color]
+    else:
+        rgb_color = _hex_to_rgb(color)
+        if rgb_color is None:
+            msg = "'color' must be a recognised name or hex value such as #RRGGBB"
+            raise ValueError(msg)
+        box_rgb = rgb_color
+        text_rgb = _get_contrasting_rgb(rgb_color)
+
+    colors = [_rgb_to_bgr(item) for item in (box_rgb, text_rgb)]
     color, color_text = colors
 
-    _cv2.rectangle(image, (left, top), (right, bottom), color, 2)
+    image_height, image_width, _ = image.shape
+
+    # The line width of the box and the font size are calculated based on the image size and the vis_size
+    # vis_size is set by the user (0=XS, 1=S, 2=M, 3=L, 4=XL)
+    if size == 0: # XS
+        width_factor = 1 / 1200
+        font_factor = width_factor * 10
+    elif size == 1: # S
+        width_factor = 1 / 600
+        font_factor = width_factor * 8
+    elif size == 2: # M
+        width_factor = 1 / 300
+        font_factor = width_factor * 6
+    elif size == 3: # L
+        width_factor = 1 / 150
+        font_factor = width_factor * 4
+    elif size == 4: # XL
+        width_factor = 1 / 75
+        font_factor = width_factor * 2
+
+    # calculate the line width of the box
+    line_width_box = max(round(min(image_height, image_width) * width_factor), 1)
+    _cv2.rectangle(image, (left, top), (right, bottom), color, line_width_box)
 
     if label:
-        _, image_width, _ = image.shape
+
+        # caluculate the font size
+        font_height = round(min(image_height, image_width) * font_factor)
+        font = ImageFont.truetype(_FONT_PATH, font_height)
+
+        label_image =  _get_label_image(label, color_text, color, font)
 
-        label_image =  _get_label_image(label, color_text, color)
         label_height, label_width, _ = label_image.shape
 
         rectangle_height, rectangle_width = 1 + label_height, 1 + label_width