import sys
import csv
import hmm
import bayes_net
from collections import OrderedDict
import song_collections
models = [("Bayes Net", bayes_net.score)] + [("{0} Order HMM".format(i), hmm.get_scorer(i)) for i in range(2, 7)]
models = OrderedDict(models)
writer = csv.writer(sys.stdout)
header_row = ["File", "Type"]
for label, model in models.items():
header_row.append(label)
header_row.append(r"% diff from random")
writer.writerow(header_row)
random_model_values = []
for i in range(len(models)):
random_model_values.append(list())
for song in song_collections.random_songs:
row = [song, "random"]
index = 0
for model in models.values():
value = model(song)
random_model_values[index].append(value)
files_to_test = []
test_path = os.path.join("data", "Random Songs")
for root, dirs, files in os.walk(test_path):
for name in [a_file for a_file in files if a_file[-4:] == ".mid"]:
files_to_test.append(os.path.join(root,name))
log_path = "log"
model_scores = OrderedDict()
for hmm_depth in range(2,7):
model_label = "HMM: %d Depth" % (hmm_depth,)
print model_label
cache = dict()
model = hmm.get_scorer(hmm_depth, cache)
scores_for_model = utils.score_files_with_model(files_to_test, model)
num_scores = len(scores_for_model)
total_scores = sum([score for score in scores_for_model.values() if score is not None])
average_score = float(total_scores) / num_scores
log_name = "HMM_" + str(hmm_depth) + "_random.txt"
log_full_path = os.path.join(log_path, log_name);
f = open(log_full_path, 'w')
index = 1
for score in scores_for_model.values():
if score is not None:
print >> f, "%d %f" % (index, score)
index += 1
f.close()