Import libraries and dataset.¶

In [61]:
from collections import Counter
from sklearn.cluster import KMeans
from matplotlib import colors
import matplotlib.pyplot as plt
import numpy as np
import cv2
from PIL import Image
import requests
import webcolors
import pandas as pd
pd.options.display.max_columns = 7
In [2]:
lea_card_data = pd.read_csv("C:\\Users\\chris\\Desktop\\lea_card_data.csv")

Create subset with art/artist data.¶

In [62]:
df = pd.concat([lea_card_data.name_x, lea_card_data.types, lea_card_data.colorIdentity, lea_card_data.artist_x,
                lea_card_data.setCode, lea_card_data.released_at, lea_card_data.scryfallid, lea_card_data.artlink], axis=1)
df
Out[62]:
name_x types colorIdentity ... released_at scryfallid artlink
0 Animate Wall ['Enchantment'] ['W'] ... 1993-08-05 d5c83259-9b90-47c2-b48e-c7d78519e792 https://c1.scryfall.com/file/scryfall-cards/ar...
1 Armageddon ['Sorcery'] ['W'] ... 1993-08-05 5b6ddce7-b9c5-431d-a0b0-46d4aa93cbcb https://c1.scryfall.com/file/scryfall-cards/ar...
2 Balance ['Sorcery'] ['W'] ... 1993-08-05 6f9ea46a-411f-40ce-a873-a905180093f4 https://c1.scryfall.com/file/scryfall-cards/ar...
3 Benalish Hero ['Creature'] ['W'] ... 1993-08-05 11600105-56c6-4073-a4a6-8469030b39c9 https://c1.scryfall.com/file/scryfall-cards/ar...
4 Black Ward ['Enchantment'] ['W'] ... 1993-08-05 15967a39-303f-457d-bcde-51837c8d63e1 https://c1.scryfall.com/file/scryfall-cards/ar...
... ... ... ... ... ... ... ...
290 Swamp ['Land'] ['B'] ... 1993-08-05 edeba4e1-9012-4962-8797-3a1f6a660952 https://c1.scryfall.com/file/scryfall-cards/ar...
291 Mountain ['Land'] ['R'] ... 1993-08-05 eace2c85-976c-425e-9800-5a6ccbd91b56 https://c1.scryfall.com/file/scryfall-cards/ar...
292 Mountain ['Land'] ['R'] ... 1993-08-05 1a96315d-edc6-4d68-a260-d14f61281dc1 https://c1.scryfall.com/file/scryfall-cards/ar...
293 Forest ['Land'] ['G'] ... 1993-08-05 6f1c8cb0-38eb-408b-94e8-16db83999b3b https://c1.scryfall.com/file/scryfall-cards/ar...
294 Forest ['Land'] ['G'] ... 1993-08-05 f20c89d9-71c9-45f5-a9cb-6e253b0a7cca https://c1.scryfall.com/file/scryfall-cards/ar...

295 rows × 8 columns

Functions to return most similar color for a given hex code.¶

In [5]:
def get_approx_color(hex_color):
    orig = webcolors.hex_to_rgb(hex_color)
    similarity = {}
    for hex_code, color_name in webcolors.CSS3_HEX_TO_NAMES.items():
        approx = webcolors.hex_to_rgb(hex_code)
        similarity[color_name] = sum(np.subtract(orig, approx) ** 2)
    return min(similarity, key=similarity.get)

def get_color_name(hex_color):
    try:
        return webcolors.hex_to_name(hex_color)
    except ValueError:
        return get_approx_color(hex_color)

Functions to return most prominent colors in an image.¶

In [46]:
def preprocess(raw):
    image = cv2.resize(raw, (900, 600), interpolation = cv2.INTER_AREA)                                          
    image = image.reshape(image.shape[0]*image.shape[1], 3)
    return image

def rgb_to_hex(rgb_color):
    hex_color = "#"
    for i in rgb_color:
        hex_color += ("{:02x}".format(int(i)))
    return hex_color

def analyze(img):
    mod_image = preprocess(img)
    
    clf = KMeans(n_clusters = 5)
    color_labels = clf.fit_predict(mod_image)
    center_colors = clf.cluster_centers_
    counts = Counter(color_labels)
    ordered_colors = [center_colors[i] for i in counts.keys()]
    hex_colors = [rgb_to_hex(ordered_colors[i]) for i in counts.keys()]

    #plt.figure(figsize = (12, 8))
    #plt.pie(counts.values(), labels = hex_colors, colors = hex_colors)
    #plt.savefig("my_pie.png")
    
    total = 0
    for i in list(counts.values()):
        total += i
    tuple_list = []
    color_list = []
    for index,color in enumerate(hex_colors):
        tuple_list.append(tuple([get_color_name(color), color, round((list(counts.values())[index]/total)*100)]))
    tuple_list.sort(key = lambda x: x[2], reverse = True)
    for tup in tuple_list:
        for j in tup:
            color_list.append(j)
    #print(color_list)
    return color_list

Grab and convert image, analyze, and add results to list. (~25 min/set)¶

In [50]:
url_list = df.artlink.tolist()
analysis_list = []
for url in url_list:
    im = Image.open(requests.get(url, stream=True).raw).convert('RGB')
    cv2_image = np.array(im)
    analysis_list.append(analyze(cv2_image))

Use results to create dataframe.¶

In [63]:
df_analysis = pd.DataFrame(analysis_list, columns=["col_string1", "hex1", "col_prop1",
                                                   "col_string2", "hex2", "col_prop2",
                                                   "col_string3", "hex3", "col_prop3",
                                                   "col_string4", "hex4", "col_prop4",
                                                   "col_string5", "hex5", "col_prop5",])
df_analysis
Out[63]:
col_string1 hex1 col_prop1 ... col_string5 hex5 col_prop5
0 silver #babbbb 25 ... darkslategray #4b4951 14
1 dimgray #7c676c 37 ... darkslategray #332631 3
2 sienna #9f7c3c 26 ... gray #898282 14
3 rosybrown #af9380 33 ... dimgray #6e5a54 7
4 dimgray #655353 31 ... dimgray #786663 8
... ... ... ... ... ... ... ...
290 silver #c6b9aa 31 ... dimgray #736a5e 9
291 dimgray #5d575a 32 ... silver #c3c7be 3
292 cadetblue #5e95ac 29 ... darkslategray #24272b 9
293 teal #1b6474 26 ... steelblue #367b89 13
294 darkslategray #3b403e 34 ... dimgray #535b5c 3

295 rows × 15 columns

Combine dataframes horizontally. Save and export to CSV.¶

In [52]:
merged_df = pd.concat([df, df_analysis], axis=1)
In [54]:
merged_df.to_csv("C:\\Users\\chris\\Desktop\\lea_color_data1.csv", index=False)