Add low precision color detection (#17)

* Add low precision color detection

* Add blank line at end of file

* Remove tab

* Remove tab

* Remove useless comment

Co-authored-by: Erick <74315338+erickahmed@users.noreply.github.com>

* Update vision-env/src/vision-test.py

Co-authored-by: Erick <74315338+erickahmed@users.noreply.github.com>

* Update vision-env/src/vision-test.py

Co-authored-by: Erick <74315338+erickahmed@users.noreply.github.com>

* Update vision-env/src/vision-test.py

Co-authored-by: Erick <74315338+erickahmed@users.noreply.github.com>

Co-authored-by: Erick <74315338+erickahmed@users.noreply.github.com>

vision-env/src/colors_detection.py

@@ -0,0 +1,69 @@
+from scipy.spatial import KDTree
+from webcolors import (
+    hex_to_rgb,
+)
+
+COLORS = {
+    "aqua": ["#00ffff", (0,255,255)],
+    "black": ["#000000", (0,0,0)],
+    "blue": ["#0000ff", (0,0,255)],
+    "fuchsia": ["#ff00ff", (255,0,255)],
+    "green": ["#008000", (0,128,0)],
+    "gray": ["#808080", (128,128,128)],
+    "lime": ["#00ff00", (0,255,0)],
+    "olive": ["#808000", (128,128,0)],
+    "purple": ["#800080", (128,0,128)],
+    "red": ["#ff0000", (255,0,0)],
+    "silver": ["#c0c0c0", (192,192,192)],
+    "teal": ["#008080", (0,128,128)],
+    "white": ["#ffffff", (255,255,255)],
+    "yellow": ["#ffff00", (255,255,0)],
+    "beige": ["#f5f5dc", (245,245,220)],
+    "brown": ["#a52a2a", (165,42,42)],
+    "gold": ["#ffd700", (255,215,0)],
+    "pink": ["#ffc0cb", (255,192,203)],
+    "lavender": ["#e6e6fa", (230,230,250)],
+    "turquoise": ["#40e0d0", (64,224,208)],
+    "violet": ["#ee82ee", (238,130,238)],
+    "orange": ["#ffa500", (255,165,0)],
+    "lightblue": ["#add8e6", (173,216,230)]
+}
+
+COLORS_HEX = {
+    "aqua": "#00ffff",
+    "black": "#000000",
+    "blue": "#0000ff",
+    "fuchsia": "#ff00ff",
+    "green": "#008000",
+    "gray": "#808080",
+    "lime": "#00ff00",
+    "olive": "#808000",
+    "purple": "#800080",
+    "red": "#ff0000",
+    "silver": "#c0c0c0",
+    "teal": "#008080",
+    "white": "#ffffff",
+    "yellow": "#ffff00",
+    "beige": "#f5f5dc",
+    "brown": "#a52a2a",
+    "gold": "#ffd700",
+    "pink": "#ffc0cb",
+    "lavender": "#e6e6fa",
+    "turquoise": "#40e0d0",
+    "violet": "#ee82ee",
+    "orange": "#ffa500",
+    "lightblue": "#add8e6",
+}
+
+def convert_rgb_to_names(rgb_tuple):
+    # A dictionary of all the color names and their respective hex in color_db
+    color_db = COLORS_HEX
+    names = []
+    rgb_values = []
+    for color_name, color_hex in color_db.items():
+        names.append(color_name)
+        rgb_values.append(hex_to_rgb(color_hex))
+    
+    kdt_db = KDTree(rgb_values)
+    distance, index = kdt_db.query(rgb_tuple)
+    return str(names[index])

vision-env/src/vision-test.py

@@ -1,16 +1,22 @@
-from genericpath import isfile
 import io
 import os
 import sys
 import numpy as np
 import cv2
+import math
+import colors_detection
 import json
+from genericpath import isfile
 from pyautogui import size
 from google.cloud import vision
 
 # Minimum score required for labels detection
 MIN_SCORE_REQUIRED = 0.80
 
+# Minimum pixel_fraction required for color detection.
+# Actually unused but useful for filtering colors.
+MIN_PFRACTION_REQUIRED = 0.02
+
 def shape_selection(event, x, y, flags, param):
     # grab references to the global variables
     global ref_point, crop
@@ -138,15 +144,45 @@ client = vision.ImageAnnotatorClient()
 image = vision.Image(content=content)
 
 # Performs label detection on the image file
-response = client.label_detection(image=image)
+label_response = client.label_detection(image=image)
-labels = response.label_annotations
+labels = label_response.label_annotations
+
+# Performs image_properties detection on the image file
+colors_response = client.image_properties(image=image)
+colors = colors_response.image_properties_annotation.dominant_colors.colors
 
 # Labels sorted by attribute score
 sorted_labels = sorted(labels, key=lambda x:x.score, reverse=True)
-
 # Labels filtered by attribute score
 filtered_labels = filter(lambda x:x.score > MIN_SCORE_REQUIRED, sorted_labels)
 
-print('Labels:')
+# Labels sorted by attribute score
+sorted_colors = sorted(colors, key=lambda x:x.pixel_fraction, reverse=True)
+# Colors filtered by attribute pixel_fraction
+#filtered_colors = filter(lambda x:x.pixel_fraction > MIN_PFRACTION_REQUIRED, sorted_colors)
+# Actually no filtering policy
+filtered_colors = sorted_colors
+
+print('\nLabels:')
 for label in filtered_labels:
-    print(label.description + ' ---> ' + str(label.score))
+    print('--- Label: ' + label.description + ' ---> ' + str(math.trunc(label.score*100)) + '%' +
+            '\n')
+
+print('\nColors:')
+for color_info in filtered_colors:
+    color = color_info.color
+
+    red = math.trunc(color.red)
+    green = math.trunc(color.green)
+    blue = math.trunc(color.blue)
+
+    rgb_triplet = (red, green, blue)
+    rgb_triplet_str = (str(red) + '%, ', str(green) + '%, ', str(blue) + '%')
+    color_name = colors_detection.convert_rgb_to_names(rgb_triplet)
+
+    # uncomment for verbose output
+    #print(str(color_info.pixel_fraction) + ' == ' + str(math.trunc(color_info.pixel_fraction*100)) + '%' + '   score: ' + str(color_info.score) + '---> ' + color_name)
+    print('--- Pixel Fraction: ' + str(round(color_info.pixel_fraction*100, 3)) + '%' + 
+            '\tScore: ' + str(round(color_info.score, 5)) + 
+            '\tColor Name: ' + color_name + 
+            '\n')