In [1]:

    

import claudio
import librosa
import matplotlib.pyplot as plt
import mir_eval
import numpy as np
import os
import pandas
import scipy.signal as sig
import IPython.display as display
import subprocess

%matplotlib inline

import seaborn
import mpld3

seaborn.set()
np.set_printoptions(precision=4, suppress=True)
mpld3.enable_notebook()


    

In [2]:

    

import minst.taxonomy as tax
import minst.sources.uiowa as uiowa
import minst.sources.rwc as rwc
import minst.signal as S


    

In [3]:

    

res = rwc.collect("/Volumes/tiny/RWC Instruments/", fext='*.ogg')
res.shape


    

    




/usr/local/lib/python3.5/site-packages/IPython/core/formatters.py:92: DeprecationWarning: DisplayFormatter._ipython_display_formatter_default is deprecated: use @default decorator instead.
  def _ipython_display_formatter_default(self):
/usr/local/lib/python3.5/site-packages/IPython/core/formatters.py:669: DeprecationWarning: PlainTextFormatter._singleton_printers_default is deprecated: use @default decorator instead.
  def _singleton_printers_default(self):






    
Out[3]:





(3544, 4)

In [4]:

    

classmap = tax.InstrumentClassMap()
dset = res[res["instrument"].isin(classmap.allnames)]
dset.shape


    

    
Out[4]:





(612, 4)

In [6]:

    

dset.to_json("/Volumes/tiny/rwc_index.json")


    

In [26]:

    

res = rwc.collect("minst/tests/dummy_rwc/")


    

In [72]:

    

res


    

    
Out[72]:






  

    

      

      
audio_file
      
dataset
      
dynamic
      
instrument
      
note
    
  
  

    

      
uiowaf22b2666
      
minst/tests/dummy_uiowa/theremin.music.uiowa.e...
      
uiowa
      
pp
      
Tuba
      
C2B2
    
  

In [123]:

    

x, fs = claudio.read(res.iloc[0].audio_file, samplerate=22050, 
                     channels=1, bytedepth=2)


    

In [215]:

    

filename = res.loc['uiowa880e2815'].audio_file
x, fs = claudio.read(filename, 
                     samplerate=22050, channels=1, bytedepth=2)
time_range = np.arange(len(x)) / fs
display.Audio(data=x.flatten(), rate=fs)


    

    
Out[215]:





                
              

In [216]:

    

x.shape


    

    
Out[216]:





(1319989, 1)

In [217]:

    

log_env = 10*np.log10(10.**-4.5 + np.power(x.flatten()[:], 2.0))


    

In [218]:

    

w_n = np.hanning(100)
w_n /= w_n.sum()
log_env_lpf = sig.filtfilt(w_n, np.ones(1), log_env)


    

In [219]:

    

fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(14, 5))
n_hop = 100
ax.plot(time_range[::n_hop], log_env_lpf[::n_hop])


    

    
Out[219]:





[<matplotlib.lines.Line2D at 0x161eb22e8>]






    

In [18]:

    

plt.plot(canny(100, 3.5, 1))


    

    
Out[18]:





[<matplotlib.lines.Line2D at 0x10c367c88>]






    

In [221]:

    

kernel = canny(100, 3.5, 1)
kernel /= 2*np.abs(kernel).sum()
onsets_forward = sig.lfilter(kernel, np.ones(1), log_env_lpf[::n_hop] - log_env_lpf.min(), axis=0)
onsets_lenv = sig.lfilter(-kernel, np.ones(1), onsets_forward[::-1], axis=0)[::-1]
onsets_pos = onsets_forward * (onsets_forward>0)
onsets_neg = onsets_forward - onsets_neg
fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(14, 5))
ax.plot(time_range[::n_hop], )
peak_idx = librosa.util.peak_pick(onsets_forward * (onsets_forward>0),
                                  500, 500, 10, 10, 0.025, 5)
peak_times = librosa.frames_to_time(peak_idx, hop_length=n_hop)
ax.vlines(peak_times, ymin=0, ymax=onsets_forward.max())


    

    
Out[221]:





<matplotlib.collections.LineCollection at 0x15d832940>






    

In [252]:

    

lcqt.shape


    

    
Out[252]:





(192, 1290)

In [223]:

    

librosa.onset.onset_detect(x.flatten(), sr=fs)


    

    
Out[223]:





array([  43,   53,   72,   80,  118,  128,  137,  155,  161,  238,  281,
        285,  295,  311,  329,  334,  381,  396,  413,  456,  487,  500,
        548,  555,  583,  588,  594,  601,  628,  670,  682,  704,  715,
        743,  784,  790,  807,  829,  880,  883,  889,  903,  911,  983,
       1010, 1025, 1082, 1105, 1125, 1155, 1188, 1210, 1242, 1250, 1293,
       1306, 1333, 1339, 1358, 1410, 1429, 1469, 1510, 1549, 1559, 1613,
       1653, 1660, 1686, 1692, 1696, 1738, 1749, 1770, 1773, 1782, 1795,
       1832, 1874, 1890, 1892, 1934, 1943, 1977, 1987, 2056, 2150, 2191,
       2212, 2243, 2258, 2281, 2366, 2371, 2415, 2420, 2424, 2453, 2488,
       2492])

In [212]:

    

librosa.util.peak_pick?


    

In [52]:

    

bins, edges = np.histogram(10*np.log(10.**-7 + np.power(x.flatten()[:], 2.0)), bins=25)
plt.bar(edges[:-1], bins, width=np.abs(np.diff(edges))[0])


    

    
Out[52]:





<Container object of 25 artists>






    

In [249]:

    

hop_length = 1024
x_noise = x + np.random.normal(scale=10.**-3, size=x.shape)
cqt = librosa.cqt(x_noise.flatten(), 
                  sr=fs, hop_length=hop_length, fmin=27.5,
                  n_bins=24*8, bins_per_octave=24, tuning=0, sparsity=0,
                  real=False, norm=1)
cqt = np.abs(cqt)


    

In [277]:

    

# lcqt = cqt - np.median(cqt, axis=1, keepdims=True)
# lcqt /= np.std(cqt, axis=1, keepdims=True)
lcqt = np.log1p(5000*cqt)


    

In [203]:

    

lcqt = librosa.logamplitude(np.power(cqt - np.median(cqt, axis=1, keepdims=True), 2.0),
                            ref_power=np.max, top_db=65)


    

In [285]:

    

plt.plot(np.median(lcqt, axis=1))
plt.plot(90. / librosa.cqt_frequencies(192, 27.5, 24, 0))


    

    
Out[285]:





[<matplotlib.lines.Line2D at 0x16b895be0>]






    

In [311]:

    

weights = (90. / librosa.cqt_frequencies(192, 27.5, 24, 0)).reshape(-1, 1)
lcqt2 = lcqt / weights
lcqt2 -= lcqt2.mean(axis=1, keepdims=True)
oset = sig.lfilter(canny(25, 3.5, 1), np.ones(1), lcqt2)
plt.imshow(oset * (oset > 0), 
           interpolation='nearest', aspect='auto', 
           origin='lower', cmap=plt.cm.Reds)


    

    
Out[311]:





<matplotlib.image.AxesImage at 0x16624c080>






    

In [329]:

    

fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(14, 6))
nov_curve = (oset * (oset > 0)).std(axis=0)
ax.plot(nov_curve)
idx = librosa.onset.onset_detect(onset_envelope=nov_curve, delta=0.05, wait=50)
ax.vlines(idx, ymin=0, ymax=23, 
           alpha=0.5, linewidth=4)


    

    
Out[329]:





<matplotlib.collections.LineCollection at 0x166f29fd0>






    

In [326]:

    

nov_curve[idx]


    

    
Out[326]:





array([ 0.5818,  0.5238,  0.6055,  0.6243,  0.6321,  0.6607,  0.7517,
        0.6024,  0.7787,  0.8289])

In [276]:

    

note_on = np.zeros([15, 75])
note_on[7:10, 25:] = 10.0
note_on -= note_on.mean()
enhance = sig.convolve2d(lcqt, note_on, mode='same')
plt.imshow(enhance * (enhance > 0), 
           interpolation='nearest', aspect='auto', 
           origin='lower', cmap=plt.cm.Reds)


    

    
Out[276]:





<matplotlib.image.AxesImage at 0x15a632048>






    

In [261]:

    

plt.plot(np.median(lcqt - lcqt_mf, axis=1))


    

    
Out[261]:





[<matplotlib.lines.Line2D at 0x169dc4c88>]






    

In [206]:

    

librosa.display.specshow(lcqt, fmin=27.5, hop_length=hop_length,
                         y_axis='cqt_hz', x_axis='time')


    

    
Out[206]:





<matplotlib.image.AxesImage at 0x154a3c7f0>






    

In [21]:

    

time_points = librosa.frames_to_time(np.arange(cqt.shape[1]), sr=fs, 
                                     hop_length=hop_length)


    

In [22]:

    

tlen = 1000
env = lcqt.mean(axis=0)
plt.plot(time_points[:tlen], env[:tlen])


    

    
Out[22]:





[<matplotlib.lines.Line2D at 0x108e5a828>]






    

In [132]:

    

bins, edges = np.histogram(np.minimum(env, env.mean()*np.ones_like(env)), 
                           bins=25, range=(0, env.mean()))


    

In [133]:

    

plt.bar(edges[:-1], bins, width=np.abs(np.diff(edges))[0])


    

    
Out[133]:





<Container object of 25 artists>






    

In [6]:

    

def canny(alpha, beta, sig=2.0):
    n = np.linspace(-beta, beta, alpha)
    kernel = (-n / np.power(sig, 2.)) * np.exp(-(n ** 2.0) / (2.0 * (sig ** 2.0)))
    return kernel


    

In [17]:

    

plt.plot(canny(15, 3.5, 1))


    

    
Out[17]:





[<matplotlib.lines.Line2D at 0x10fe50b38>]






    

In [36]:

    

cfilt = canny(15, 5, 1)
onset_lcqt = sig.lfilter(cfilt, np.ones(1), lcqt, axis=1)
plt.imshow(np.abs(onset_lcqt), interpolation='nearest', aspect='auto', 
           origin='lower', cmap=plt.cm.Reds)


    

    
Out[36]:





<matplotlib.image.AxesImage at 0x1112fb668>






    

In [37]:

    

tlen = 1000
# env2 = onset_lcqt.mean(axis=0)
env2 = np.max(onset_lcqt, axis=0)
plt.plot(time_points[:tlen], env2[:tlen])


    

    
Out[37]:





[<matplotlib.lines.Line2D at 0x111748dd8>]






    

In [27]:

    

lr_onset = librosa.onset.onset_strength(S=lcqt, aggregate=np.median)


    

In [28]:

    

fig, axes = plt.subplots(nrows=1, ncols=1, figsize=(12, 8))
axes.plot(time_points[:tlen], lr_onset[:tlen])
# axes[1].plot(time_points[:tlen], env2[:tlen] * (env2[:tlen] > 0))


    

    
Out[28]:





[<matplotlib.lines.Line2D at 0x10c6d34a8>]






    

In [94]:

    

fs


    

    
Out[94]:





22050.0

In [140]:

    

onset_idxs = librosa.onset.onset_detect(onset_envelope=lr_onset)


    

In [149]:

    

start_times = librosa.frames_to_time(onset_idxs, hop_length=hop_length)


    

In [152]:

    

fig, axes = plt.subplots(nrows=1, ncols=1, figsize=(12, 5))
trange = np.arange(len(x)) / fs
nhop = 1
axes.plot(trange[::nhop], x.flatten()[::nhop])
max_val = np.abs(x).max()
axes.vlines(start_times, ymax=max_val, ymin=-max_val, colors='black')
display.Audio(data=x.flatten(), rate=fs)


    

    
Out[152]:





                
              






    

In [224]:

    

librosa.util.peak_pick?


    

In [242]:

    

def segment_one(filename):
    x, fs = claudio.read(filename, samplerate=22050, 
                         channels=1, bytedepth=2)
    time_range = np.arange(len(x)) / fs
    
    log_env = 10*np.log10(10.**-4.5 + np.power(x.flatten()[:], 2.0))
    w_n = np.hanning(100)
    w_n /= w_n.sum()
    log_env_lpf = sig.filtfilt(w_n, np.ones(1), log_env)
    
    n_hop = 100
    kernel = canny(100, 3.5, 1)
    kernel /= np.abs(kernel).sum()
    onsets_forward = sig.lfilter(kernel, np.ones(1), log_env_lpf[::n_hop] - log_env_lpf.min(), axis=0)
    onsets_lenv = sig.lfilter(-kernel, np.ones(1), onsets_forward[::-1], axis=0)[::-1]
    onsets_pos = onsets_forward * (onsets_forward>0)
    onsets_neg = onsets_forward - onsets_pos

    fig, axes = plt.subplots(nrows=3, ncols=1, figsize=(14, 12))
    axes[0].plot(time_range[::n_hop], x.flatten()[::n_hop])
    peak_idx = librosa.util.peak_pick(onsets_forward * (onsets_forward>0),
                                      pre_max=500, post_max=500, pre_avg=10, 
                                      post_avg=10, delta=0.025, wait=100)
    peak_times = librosa.frames_to_time(peak_idx, hop_length=n_hop)
    axes[1].plot(time_range[::n_hop], log_env_lpf[::n_hop])
    axes[2].plot(time_range[::n_hop], onsets_pos)
    for ax, yvals in zip(axes, [(x.min(), x.max()), (log_env_lpf.min(), 0), (0, onsets_forward.max())]):
        ax.vlines(peak_times, ymin=yvals[0], ymax=yvals[1])
    
    return display.Audio(data=x.flatten(), rate=fs)


    

In [247]:

    

filename = res[res.instrument == "Trumpet"].iloc[1].audio_file
print(filename)
segment_one(filename)


    

    




/Volumes/camel/instruments/uiowa/theremin.music.uiowa.edu/sound files/MIS/Brass/Bbtrumpet/Trumpet.novib.ff.C5B5.aiff






    
Out[247]:





                
              






    

In [337]:

    

def segment_one(filename):
    x, fs = claudio.read(filename, samplerate=22050, 
                     channels=1, bytedepth=2)
    time_range = np.arange(len(x)) / fs
    
    hop_length = 1024
    x_noise = x + np.random.normal(scale=10.**-3, size=x.shape)
    cqt = librosa.cqt(x_noise.flatten(), 
                      sr=fs, hop_length=hop_length, fmin=27.5,
                      n_bins=24*8, bins_per_octave=24, tuning=0, 
                      sparsity=0, real=False, norm=1)
    cqt = np.abs(cqt)
    lcqt = np.log1p(5000*cqt)
    weights = (96. / librosa.cqt_frequencies(192, 27.5, 24, 0)).reshape(-1, 1)
    lcqt2 = lcqt / weights
    lcqt2 -= lcqt2.mean(axis=1, keepdims=True)
    oset = sig.lfilter(canny(25, 3.5, 1), np.ones(1), lcqt2)
    n_hop = 100
    
    fig, axes = plt.subplots(nrows=3, ncols=1, figsize=(14, 12))
    axes[0].plot(time_range[::n_hop], x.flatten()[::n_hop])
    axes[1].imshow(oset * (oset > 0), 
                   interpolation='nearest', aspect='auto', 
                   # (left, right, bottom, top)
                   extent=(0, time_range.max(), 0, 192),
                   origin='lower', cmap=plt.cm.Reds)
    

    nov_curve = (oset * (oset > 0)).std(axis=0)
    axes[2].plot(librosa.frames_to_time(np.arange(len(nov_curve)), 
                                        hop_length=hop_length), 
                 nov_curve)
    peak_idx = librosa.onset.onset_detect(onset_envelope=nov_curve, delta=0.05, wait=50)
    peak_times = librosa.frames_to_time(peak_idx, hop_length=hop_length)
    lims = [(1.2*x.min(), 1.2*x.max()), (0, 192), (nov_curve.min(), nov_curve.max())]
    for ax, yvals in zip(axes, lims):
        ax.vlines(peak_times, ymin=yvals[0], ymax=yvals[1], 
                  alpha=0.5, linewidth=3)
    
#     ax.vlines(idx, ymin=0, ymax=23, 
#                alpha=0.5, linewidth=4)


    

In [338]:

    

filename = res[res.instrument == "Trumpet"].iloc[1].audio_file
print(filename)
segment_one(filename)


    

    




/Volumes/camel/instruments/uiowa/theremin.music.uiowa.edu/sound files/MIS/Brass/Bbtrumpet/Trumpet.novib.ff.C5B5.aiff






    

In [11]:

    

def segment_one(filename, pre_max=0, post_max=1, pre_avg=0, 
                post_avg=1, delta=0.05, wait=50):
    x, fs = claudio.read(filename, samplerate=22050, 
                     channels=1, bytedepth=2)
    time_range = np.arange(len(x)) / fs
    
    hop_length = 1024
    x_noise = x + np.random.normal(scale=10.**-3, size=x.shape)
    cqt = librosa.cqt(x_noise.flatten(), 
                      sr=fs, hop_length=hop_length, fmin=27.5,
                      n_bins=24*8, bins_per_octave=24, tuning=0, 
                      sparsity=0, real=False, norm=1)
    cqt = np.abs(cqt)
    lcqt = np.log1p(5000*cqt)
    return lcqt
    c_n = canny(51, 3.5, 1)
    onset_strength = sig.lfilter(c_n, np.ones(1), lcqt, axis=1).mean(axis=0)
    
    
    n_hop = 100
    fig, axes = plt.subplots(nrows=3, ncols=1, figsize=(14, 12))
    axes[0].plot(time_range[::n_hop], x.flatten()[::n_hop])
    axes[1].imshow(lcqt, 
                   interpolation='nearest', aspect='auto', 
                   # (left, right, bottom, top)
                   extent=(0, time_range.max(), 0, 192),
                   origin='lower', cmap=plt.cm.Blues)
    
    axes[2].plot(librosa.frames_to_time(np.arange(len(onset_strength)), 
                                        hop_length=hop_length), 
                 onset_strength)
    peak_idx = librosa.onset.onset_detect(
        onset_envelope=onset_strength, delta=delta, wait=wait,
        )
    peak_times = librosa.frames_to_time(peak_idx, hop_length=hop_length)
    lims = [(1.2*x.min(), 1.2*x.max()), (0, 192), 
            (onset_strength.min(), onset_strength.max())]
    for ax, yvals in zip(axes, lims):
        ax.vlines(peak_times, ymin=yvals[0], ymax=yvals[1], 
                  alpha=0.3, linewidth=3, colors='green')
    
#     ax.vlines(idx, ymin=0, ymax=23, 
#                alpha=0.5, linewidth=4)


    

In [71]:

    

filename = res[res.instrument == "Trumpet"].iloc[1].audio_file
print(filename)
lcqt = segment_one(filename, delta=0.1, wait=100)


    

    




---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-71-9a9533c24fe3> in <module>()
----> 1 filename = res[res.instrument == "Trumpet"].iloc[1].audio_file
      2 print(filename)
      3 lcqt = segment_one(filename, delta=0.1, wait=100)

AttributeError: 'numpy.ndarray' object has no attribute 'instrument'

In [16]:

    

fig, axes = plt.subplots(nrows=2, ncols=1, figsize=(14, 8))
axes[0].imshow(lcqt, 
               interpolation='nearest', aspect='auto', 
               origin='lower', cmap=plt.cm.Blues)

c_n = canny(51, 3.5, 1)
os = sig.lfilter(c_n, np.ones(1), lcqt, axis=1)
axes[1].imshow(os * (os > 0),
               interpolation='nearest', aspect='auto', 
               origin='lower', cmap=plt.cm.Blues)


    

    
Out[16]:





<matplotlib.image.AxesImage at 0x114aef160>






    

In [22]:

    

c_n = canny(100, 3.5, 1).reshape(-1, 1)
c_n2 = c_n * np.diff(c_n, axis=0).T
plt.imshow(c_n2, interpolation='nearest', aspect='auto', 
               origin='lower', cmap=plt.cm.RdBu)


    

    
Out[22]:





<matplotlib.image.AxesImage at 0x10f581f98>






    

In [25]:

    

wtf = sig.convolve2d(lcqt, c_n2.T, mode='same')


    

In [26]:

    

fig, axes = plt.subplots(nrows=2, ncols=1, figsize=(14, 8))
axes[0].imshow(lcqt, 
               interpolation='nearest', aspect='auto', 
               origin='lower', cmap=plt.cm.Blues)

axes[1].imshow(wtf,
               interpolation='nearest', aspect='auto', 
               origin='lower', cmap=plt.cm.Blues)


    

    
Out[26]:





<matplotlib.image.AxesImage at 0x1141ff8d0>






    

In [136]:

    

def hll_one(audiofile):
    tempfile = 'temp.wav'
    claudio.sox.convert(audiofile, tempfile, samplerate=44100, channels=1, bytedepth=2)
    output_csv = 'output.csv'
    if os.path.exists(output_csv):
        os.remove(output_csv)
    args = ['/Users/ejhumphrey/hll/hll_mono', tempfile, 
            output_csv, '/Users/ejhumphrey/hll/HLL_MONO_PARAMS.csv']
    subprocess.check_output(args)
    return pandas.read_csv(output_csv, header=0)


    

In [110]:

    

classmap = tax.InstrumentClassMap()
dset = res[res["instrument"].isin(classmap.allnames)]


    

In [111]:

    

filename = dset[dset.instrument == "Flute"].iloc[1].audio_file
print(filename, dset.shape)


    

    




/Volumes/camel/instruments/uiowa/theremin.music.uiowa.edu/sound files/MIS/Woodwinds/flute/Flute.nonvib.ff.C5B5.aiff (441, 5)

In [105]:

    

data = np.asarray(hll(filename)).T
trange = data[0]/44100.
nidx = data[2] > 0.999
data[1, nidx] = 0
data[2, nidx] = 0


    

In [146]:

    

filename = dset.iloc[30].audio_file
print(filename)
time_points, freqs, amps = hll.hll(filename)

freqs = sig.medfilt(freqs, 11)
amps = sig.medfilt(amps, 11)

fig, axes = plt.subplots(nrows=4, ncols=1, figsize=(14, 12))
axes[0].plot(time_points, freqs)
axes[1].plot(time_points, amps)
voicing = amps*freqs
print(voicing.mean(), np.median(voicing))
axes[2].plot(time_points, voicing)
c_n = canny(25, 3.5, 1)
onsets = sig.lfilter(c_n, [1], voicing > .5)
axes[3].plot(time_points, onsets * (onsets > 0), 
             'green', alpha=0.5)
axes[3].plot(time_points, -onsets * (onsets < 0), 
             'black', alpha=0.5)

for ax in axes:
    ax.set_xlim(0, time_points.max())


    

    




/Volumes/camel/instruments/uiowa/theremin.music.uiowa.edu/sound files/MIS/Brass/frenchhorn/Horn.mf.C2B2.aiff
4.10658090286 3.67223671427






    

In [135]:

    

filename = dset.iloc[100].audio_file
print(filename)
x, fs = claudio.read(filename, samplerate=8000, 
                     channels=1, bytedepth=2)
trange = np.arange(len(x)) / fs

onsets, offsets = hll.segment_one(
    filename, mfilt_len=51, threshold=0.15, 
    delta=0.01, wait=500)

x_max = 1.2 * np.abs(x).max()

fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(16, 4))
ax.plot(trange, x.flatten())
ax.vlines(onsets, ymin=-x_max, ymax=x_max, 
          alpha=0.3, linewidth=3, colors='green')
ax.vlines(offsets, ymin=-x_max, ymax=x_max, 
          alpha=0.3, linewidth=3, colors='black')
ax.set_xlim(0, trange.max());


    

    




/Volumes/camel/instruments/uiowa/theremin.music.uiowa.edu/sound files/MIS/Piano_Other/guitar/Guitar.mf.sulD.C4Ab4.stereo.aif






    

In [18]:

    

res.loc['uiowa32813686'].audio_file


    

    
Out[18]:





'/Volumes/SHUTTLE/uiowa/theremin.music.uiowa.edu/sound files/MIS/Strings/cello/Cello.arco.mf.sulD.C4Bb4.ogg'

In [7]:

    

onsets, fig = S.segment(res.loc['uiowa16a92810'].audio_file, 'envelope')


    

    




[  2.3084   7.229   12.263   17.161   22.0862  27.22    32.1451  37.0159
  42.0771  47.0567  52.0227  57.0023]






    

In [5]:

    

onsets


    

    
Out[5]:






  

    

      

      
env_delta
      
env_max
      
env_mean
      
env_std
      
time
    
  
  

    

      
0
      
16.446086
      
-28.553914
      
-29.592884
      
0.506350
      
2.101406
    
    

      
1
      
18.937986
      
-26.062014
      
-27.326012
      
0.725745
      
7.024036
    
    

      
2
      
19.099639
      
-25.900361
      
-26.935300
      
1.015016
      
12.056961
    
    

      
3
      
20.696745
      
-24.303255
      
-25.949579
      
1.199132
      
16.944762
    
    

      
4
      
21.852087
      
-23.147913
      
-24.410671
      
1.251316
      
21.867392
    
    

      
5
      
19.282186
      
-25.717814
      
-26.471626
      
1.389963
      
26.981587
    
    

      
6
      
18.867614
      
-26.132386
      
-26.857720
      
0.631446
      
31.933243
    
    

      
7
      
22.776237
      
-22.223763
      
-23.750676
      
1.550207
      
36.797823
    
    

      
8
      
18.912525
      
-26.087475
      
-26.557465
      
0.374487
      
41.882993
    
    

      
9
      
15.340151
      
-29.659849
      
-30.883861
      
3.145894
      
46.770794
    
    

      
10
      
16.656226
      
-28.343774
      
-28.875129
      
0.288071
      
51.821134
    
    

      
11
      
18.819975
      
-26.180025
      
-27.258997
      
0.898514
      
56.778594
    
  

In [8]:

    

res.shape


    

    
Out[8]:





(396, 5)

In [16]:

    

res.to_json("/Users/ejhumphrey/Desktop/uiowa_index.json")


    

In [ ]: