battle-extractor2.py

import json
import cv2
import copy
from PIL import ImageColor
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from st_dbscan import ST_DBSCAN

colors = ['#a6cee3', '#1f78b4', '#b2df8a', '#33a02c', '#fb9a99', '#e31a1c', '#fdbf6f', '#ff7f00', '#cab2d6', '#6a3d9a',
          '#ffff99', '#b15928']


def cc(v):
    return v / 15000


def plot(data, labels):
    for i in range(-1, len(set(labels))):
        if i == -1:
            col = [0, 0, 0, 1]
        else:
            col = colors[i % len(colors)]

        clust = data[np.where(labels == i)]
        plt.scatter(clust[:, 0], clust[:, 1], c=[col], s=1)
    plt.show()

    return None


image = cv2.imread('LoLBaseMap3.png')

f = open('Match3.json')
data = json.load(f)
frames = data["info"]["frames"]

df = pd.DataFrame(columns=['time', 'x', 'y'])
original_times = []
currentPlayer = []
factor = 1000

events = []
for i in range(len(frames)):
    events.append(frames[i]['events'])

max_time = 0
# looop through champion kills events and add location and time info to df
for i in range(len(events)):
    for j in range(len(events[i])):
        e = events[i][j]
        if events[i][j]['type'] == "CHAMPION_KILL" and events[i][j]['killerId'] != 0:
            time = events[i][j]['timestamp'] / 1000  # convert to sec
            x = cc(events[i][j]['position']['x'])
            y = cc(events[i][j]['position']['y'])

            if time > max_time:
                max_time = time

            # print(str(time) + " " + str(x) + " " + str(y))

            new_row = {'x': x, 'y': y, 'time': time}
            df = df.append(new_row, ignore_index=True)
            original_times.append(events[i][j]['timestamp'])
            currentPlayer.append(events[i][j]['victimId'])

# RUN DBSCAN
print(df['x'])

# For float-precision issue in DB-SCAN we need to scale the data

df['x'] = df['x'].mul(factor)  # scale up x and y coord to
df['y'] = df['y'].mul(factor)

print(df['x'])

#   eps1 = maximum spatial distance
#   eps2 = maximum temporal distance
#   min_samples  = number of samples required
data = df.loc[:, ['x', 'y', 'time']].values

st_dbscan = ST_DBSCAN(eps1=90, eps2=104, min_samples=3)
st_dbscan.fit(data)

# plot(data[:,1:], st_dbscan.labels)
print("CLUSTER LABELS")
print(st_dbscan.labels)

# print(df)

# normalize time and coords to [0...1]
df['time'] = df['time'].div(max_time)
df['x'] = df['x'].div(factor)  # scale up x and y coord to
df['y'] = df['y'].div(factor)

# print(df_copy)


# mark locations on map
height = image.shape[0]
factor = height / 15000

i = 0

result = []

for index, row in df.iterrows():
    x = int((row['x'] * 15000) * factor)
    y = int((15000 - row['y'] * 15000) * factor)

    P = (x, y)

    # print(df_copy.at[index, 'time'])

    color = (0, 0, 0)
    if st_dbscan.labels[i] >= 0:
        print(str(st_dbscan.labels[i]) + "  " + str(row['time'] * max_time))

        obj = {'label': int(st_dbscan.labels[i]), 'timestamp': original_times[i], 'player': 'Player' + str(currentPlayer[i])}

        result.append(obj)

        color = ImageColor.getcolor(colors[st_dbscan.labels[i]], "RGB")  # BGR
        l = list(color)
        tmp = l[0]
        l[0] = l[2]
        l[2] = tmp
        color = tuple(l)

    image = cv2.circle(image, P, radius=7, color=color, thickness=-1)
    image = cv2.circle(image, P, radius=4, color=(255 * row['time'], 255 * row['time'], 255 * row['time']),
                       thickness=-1)
    i += 1

# print(json)

jsonString = json.dumps(result, indent=2)
jsonFile = open("Results/dbscan.json", "w")
jsonFile.write(jsonString)
jsonFile.close()

cv2.imshow('image', image)  # show image and wait for keypress
cv2.waitKey(0)