Added automatic background selection based on dominant picture color and mass conversion script

.gitignore

@@ -0,0 +1,4 @@
+*.un~
+data/input/*
+data/output/*
+__pycache__

img2img_color.py

@@ -6,7 +6,7 @@
 import cv2
 import numpy as np
 from PIL import Image, ImageDraw, ImageOps
-from utils import get_data
+from utils import get_data, get_dominant_color
 
 
 def get_args():
@@ -15,7 +15,7 @@ def get_args():
     parser.add_argument("--output", type=str, default="data/output.jpg", help="Path to output text file")
     parser.add_argument("--language", type=str, default="english")
     parser.add_argument("--mode", type=str, default="standard")
-    parser.add_argument("--background", type=str, default="black", choices=["black", "white"],
+    parser.add_argument("--background", type=str, default="auto", choices=["auto", "black", "white"],
                         help="background's color")
     parser.add_argument("--num_cols", type=int, default=300, help="number of character for output's width")
     parser.add_argument("--scale", type=int, default=2, help="upsize output")
@@ -24,15 +24,27 @@ def get_args():
 
 
 def main(opt):
-    if opt.background == "white":
-        bg_code = (255, 255, 255)
-    else:
-        bg_code = (0, 0, 0)
     char_list, font, sample_character, scale = get_data(opt.language, opt.mode)
     num_chars = len(char_list)
     num_cols = opt.num_cols
     image = cv2.imread(opt.input, cv2.IMREAD_COLOR)
     image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+
+    if opt.background == "auto":
+        dominant = get_dominant_color(image)
+        diff_white = 255 - dominant[0] + 255 - dominant[1] + 255 - dominant[2]
+        if(diff_white <= 381):
+            print("WHITE Background")
+            opt.background = "white"
+        else:
+            print("BLACK Blackground")
+            opt.background = "black"
+
+    if opt.background == "white":
+        bg_code = (255, 255, 255)
+    else:
+        bg_code = (0, 0, 0)
+
     height, width, _ = image.shape
     cell_width = width / opt.num_cols
     cell_height = scale * cell_width
@@ -43,7 +55,8 @@ def main(opt):
         cell_height = 12
         num_cols = int(width / cell_width)
         num_rows = int(height / cell_height)
-    char_width, char_height = font.getsize(sample_character)
+    font_bbox = font.getbbox(sample_character)
+    char_width, char_height = font_bbox[2] - font_bbox[0], font_bbox[3]
     out_width = char_width * num_cols
     out_height = scale * char_height * num_rows
     out_image = Image.new("RGB", (out_width, out_height), bg_code)
@@ -63,6 +76,7 @@ def main(opt):
         cropped_image = out_image.getbbox()
     out_image = out_image.crop(cropped_image)
     out_image.save(opt.output)
+    print("Finished %s" % opt.input)
 
 
 if __name__ == '__main__':

img2img_color_all.sh

@@ -0,0 +1,23 @@
+#!/bin/bash
+
+# Run img2img_color.py in a loop over INPUT directory
+
+INPUT_PATH="data/input"
+OUTPUT_PATH="data/output"
+
+INPUT=`ls $INPUT_PATH`
+OUTPUT=`ls $OUTPUT_PATH`
+
+for filename in $INPUT; do
+#    echo "$filename"
+
+    if [[ $OUTPUT != *"$filename"* ]]; then
+        echo "Converting $filename"
+        python3 img2img_color.py --input=$INPUT_PATH/$filename --output=$OUTPUT_PATH/$filename
+    fi
+done
+
+
+
+#ls data/input | awk '{ system("python3 img2img_color.py --input=data/input/" $1 " --output=data/output/" $1) }'
+#ls data/input | grep "XX" | awk '{ system("echo Processing " $1 "; python3 img2img_color.py --input=data/input/" $1 " --output=data/output/" $1 " --background=white") }'

utils.py

@@ -1,4 +1,5 @@
 import numpy as np
+import cv2
 from PIL import Image, ImageFont, ImageDraw, ImageOps
 
 
@@ -117,3 +118,14 @@ def get_data(language, mode):
         char_list = sort_chars(char_list, font, language)
 
     return char_list, font, sample_character, scale
+
+def get_dominant_color(img):
+    data = np.reshape(img, (-1,3))
+    data = np.float32(data)
+
+    criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 10, 1.0)
+    flags = cv2.KMEANS_RANDOM_CENTERS
+    compactness,labels,centers = cv2.kmeans(data,1,None,criteria,10,flags)
+
+    # print('Dominant color is: bgr({})'.format(centers[0].astype(np.int32)))
+    return tuple(centers[0].astype(np.int32))