Score data extraction from Jukebeat screenshot (ripples theme only)

extract.py

import sys
from PIL import Image
import numpy
#import colorsys

def parse(path):
    im = Image.open(path).convert('RGB')
    w, h = im.size

    score = im.crop((w*137>>8, h*35>>8, w*254>>8, h*49>>8)).resize((350, 65), Image.BILINEAR)
    score = numpy.asarray(score).astype('int16')
    score = score[:,:,0] + score[:,:,1] + score[:,:,2]
    score = ~((200 < score) & (score < 500))
    parts = []
    for i in xrange(7):
        part = score[0:40,i*50+5:i*50+45]
        #Image.fromarray((part * 255).astype('uint8')).save('tmp%d.png'%i)
        part = abs(numpy.fft.rfft2(part))
        if part[0,0] != 0: part /= part[0,0]
        parts.append(list(part[:4,:4].flatten())[1:])

    return parts

def train():
    samples = [
        '      0', 'IMG_0025.PNG',
        '  22530', 'IMG_0026.PNG',
        ' 134561', 'IMG_0027.PNG',
        ' 208539', 'IMG_0028.PNG',
        ' 335966', 'IMG_0029.PNG',
        ' 424464', 'IMG_0030.PNG',
        ' 500945', 'IMG_0031.PNG',
        ' 604589', 'IMG_0032.PNG',
        ' 712864', 'IMG_0033.PNG',
        ' 713740', 'IMG_0034.PNG',
        ' 752044', 'IMG_0035.PNG',
        ' 813754', 'IMG_0036.PNG',
        ' 897738', 'IMG_0037.PNG',
        ' 177496', 'IMG_0038.PNG',
        ' 296786', 'IMG_0040.PNG',
        ' 413323', 'IMG_0041.PNG',
        ' 521473', 'IMG_0042.PNG',
        ' 757051', 'IMG_0043.PNG',
        ' 793602', 'IMG_0044.PNG',
        ' 849461', 'IMG_0045.PNG',
        ' 973712', 'jukebeat-ripples-ex.png',
        ' 942379', 'jukebeat-ripples-ex2.png',
    ]
    keys = [(j,i) for i in (True, False) for j in ' 0123456789']
    keys.remove((' ', True))
    sums = dict((k, [0]*15) for k in keys)
    numbers = dict((k, 0) for k in keys)
    for path, actual in zip(samples[1::2], samples[0::2]):
        yellow = int(actual) >= 700000
        for i, v in enumerate(parse(path)):
            k = actual[i], yellow and actual[i] != ' '
            #if numbers[k] > 0:
            #    d = sum(abs(sums[k][j]/numbers[k] - v[j]) for j in xrange(15))
            #    print d
            numbers[k] += 1
            for j in xrange(15): sums[k][j] += v[j]

    print 'AVERAGES = ['
    for (c, yellow), n in sorted(numbers.items()):
        print '    (%r, [%s]),' % (c, ', '.join('%5.3f' % (v/n) for v in sums[c, yellow]))
    print ']'

#train()

AVERAGES = [
    (' ', [0.001, 0.001, 0.000, 0.002, 0.001, 0.001, 0.000, 0.002, 0.001, 0.001, 0.000, 0.002, 0.001, 0.001, 0.000]),
    ('0', [0.133, 0.140, 0.078, 0.206, 0.254, 0.103, 0.078, 0.173, 0.058, 0.060, 0.011, 0.047, 0.056, 0.023, 0.025]),
    ('0', [0.041, 0.009, 0.055, 0.018, 0.051, 0.005, 0.036, 0.028, 0.021, 0.012, 0.014, 0.015, 0.021, 0.016, 0.019]),
    ('1', [0.230, 0.110, 0.072, 0.113, 0.033, 0.047, 0.051, 0.086, 0.089, 0.037, 0.008, 0.018, 0.041, 0.039, 0.022]),
    ('1', [0.063, 0.016, 0.025, 0.020, 0.020, 0.014, 0.015, 0.020, 0.018, 0.014, 0.004, 0.014, 0.016, 0.010, 0.008]),
    ('2', [0.136, 0.163, 0.032, 0.190, 0.191, 0.110, 0.062, 0.129, 0.162, 0.031, 0.017, 0.056, 0.048, 0.056, 0.036]),
    ('2', [0.033, 0.019, 0.013, 0.018, 0.064, 0.013, 0.011, 0.018, 0.033, 0.018, 0.060, 0.036, 0.020, 0.038, 0.036]),
    ('3', [0.199, 0.150, 0.025, 0.130, 0.127, 0.094, 0.067, 0.205, 0.184, 0.053, 0.045, 0.031, 0.021, 0.034, 0.015]),
    ('3', [0.056, 0.035, 0.012, 0.025, 0.031, 0.021, 0.006, 0.054, 0.044, 0.024, 0.027, 0.027, 0.017, 0.020, 0.032]),
    ('4', [0.221, 0.144, 0.064, 0.182, 0.212, 0.067, 0.064, 0.069, 0.025, 0.134, 0.054, 0.033, 0.027, 0.037, 0.032]),
    ('4', [0.040, 0.025, 0.025, 0.050, 0.048, 0.028, 0.013, 0.020, 0.020, 0.017, 0.026, 0.031, 0.014, 0.024, 0.052]),
    ('5', [0.154, 0.212, 0.077, 0.084, 0.140, 0.075, 0.042, 0.343, 0.100, 0.053, 0.038, 0.034, 0.063, 0.043, 0.016]),
    ('5', [0.028, 0.021, 0.015, 0.029, 0.028, 0.015, 0.041, 0.072, 0.015, 0.009, 0.022, 0.023, 0.046, 0.018, 0.018]),
    ('6', [0.100, 0.183, 0.032, 0.148, 0.130, 0.135, 0.074, 0.300, 0.111, 0.111, 0.048, 0.035, 0.017, 0.032, 0.025]),
    ('6', [0.021, 0.020, 0.035, 0.035, 0.016, 0.015, 0.033, 0.048, 0.036, 0.022, 0.015, 0.011, 0.022, 0.022, 0.026]),
    ('7', [0.194, 0.077, 0.027, 0.236, 0.138, 0.146, 0.053, 0.218, 0.076, 0.067, 0.063, 0.091, 0.040, 0.020, 0.022]),
    ('7', [0.050, 0.016, 0.006, 0.049, 0.049, 0.008, 0.027, 0.032, 0.013, 0.016, 0.041, 0.025, 0.011, 0.011, 0.005]),
    ('8', [0.174, 0.221, 0.024, 0.202, 0.092, 0.149, 0.060, 0.193, 0.203, 0.034, 0.043, 0.104, 0.029, 0.031, 0.017]),
    ('8', [0.025, 0.026, 0.015, 0.029, 0.018, 0.014, 0.025, 0.031, 0.027, 0.027, 0.048, 0.056, 0.007, 0.026, 0.016]),
    ('9', [0.041, 0.228, 0.081, 0.170, 0.214, 0.090, 0.058, 0.228, 0.095, 0.087, 0.031, 0.097, 0.083, 0.028, 0.011]),
    ('9', [0.010, 0.018, 0.054, 0.036, 0.039, 0.018, 0.045, 0.043, 0.016, 0.014, 0.007, 0.029, 0.020, 0.018, 0.017]),
]

def detect(parts):
    score = ''
    for p in parts:
        bestc = None
        bestd = 9e9
        for c, v in AVERAGES:
            d = sum(abs(a - b) for a, b in zip(v, p))
            if d < bestd:
                bestd = d
                bestc = c
        score += bestc
    return score

print detect(parse(sys.argv[1]))