comparison trunk/experiments/scripts/cnbh-syllables/results_plotting/plot_munged_results.py @ 440:99f9bf0f7798

- AIMC format file output

 """
 plot_munged_results.py
 
 """
 
+import matplotlib as mpl
+mpl.use('PDF')
 import numpy as np
 import pylab as p
-import matplotlib as mpl
-mpl.use('PDF')
-import matplotlib.pyplot as plt
-
+#import matplotlib.pyplot as plt
 
 f=open("results_test_all.csv","r")
 results = dict()
 for line in f:
   if line[0] != "#":

     results[values[3]][values[0]][values[1]].setdefault(int(values[4]), dict())
     results[values[3]][values[0]][values[1]][int(values[4])].setdefault(int(values[5]), dict())
     results[values[3]][values[0]][values[1]][int(values[4])][int(values[5])].setdefault(int(values[6]), dict())
     if values[2] == 'clean':
        snr = 50
+       results[values[3]][values[0]][values[1]][int(values[4])][int(values[5])][int(values[6])][snr] = float(values[7])
     else:
        snr = int(values[2])
        results[values[3]][values[0]][values[1]][int(values[4])][int(values[5])][int(values[6])][snr] = float(values[7])
 #    results[values[3]].append((values[1],values[2],values[2],values[4]))
 
-ax = plt.subplot(111)
-
+ax = mpl.pyplot.subplot(111)
 train_set = 'inner'
-lines = []
-labels = []
-hmm_iterations = 2
-hmm_states = 4
-hmm_components = 4
-for feature_type in ('mfcc', 'mfcc_vtln', 'aim'):
-  for feature_subtype in results[train_set][feature_type].keys():
-    this_line = results[train_set][feature_type][feature_subtype][hmm_states][hmm_components][hmm_iterations].items()
-    this_line.sort(cmp=lambda x,y: x[0] - y[0])
-    xs, ys = zip(*this_line)
-    xs = list(xs)
-    ys = list(ys)
-    line, = ax.plot(xs,ys,'-o',linewidth=2)
-    lines.append(line)
-    labels.append(feature_type + "_" + feature_subtype)
-p.legend(lines, labels, 'upper left', shadow=True)
-p.xlabel('SNR/dB')
-p.ylabel('Recognition performance %')
-plt.savefig(output_file)
+for hmm_iterations in [2,3,15]:
+  for hmm_states in [3,4]:
+    for hmm_components in [3,4]:
+      lines = []
+      labels = []
+      ax.cla()
+      for feature_type in ('mfcc', 'mfcc_vtln', 'aim'):
+        for feature_subtype in results[train_set][feature_type].keys():
+          try:
+            this_line = results[train_set][feature_type][feature_subtype][hmm_states][hmm_components][hmm_iterations].items()
+            this_line.sort(cmp=lambda x,y: x[0] - y[0])
+            xs, ys = zip(*this_line)
+            xs = list(xs)
+            ys = list(ys)
+            line, = ax.plot(xs,ys,'-o',linewidth=2)
+            lines.append(line)
+            labels.append(feature_type + "_" + feature_subtype)
+          except KeyError:
+            print "Data not found"
+      p.legend(lines, labels, 'upper left', shadow=True)
+      p.xlabel('SNR/dB')
+      p.ylabel('Recognition performance %')
+      output_file = ("recognition_vs_snr_%diterations_%dstates_%d_components.pdf" % (hmm_iterations, hmm_states, hmm_components))
+      p.savefig(output_file)