print 'do_resample =', do_resample
else:
break
sys.argv.pop(1)
# load song
songpath = sys.argv[1]
print 'song file:',songpath
dictpath = sys.argv[2]
print 'codebook file:',dictpath
# feats
feats = FEATS.features_from_matfile(songpath,pSize=pSize,
usebars=usebars,keyInv=keyInv,
songKeyInv=songKeyInv,
positive=positive,
do_resample=do_resample)
# model
mat = scipy.io.loadmat(dictpath)
codebook = mat['codebook']
model = MODEL.Model(codebook)
# predict
best_code_per_pattern, avg_dist = model.predicts(feats)
# report distortion (per... pixel? patch point?)
print 'average distortion:', np.average(avg_dist)
# build original and encoding
patch_len = codebook.shape[1]/12
def next_track(self, auto_bar=None):
"""
Returns features for a random matlab file, or None if some error.
If only one iter over the data, raise StopIteration at the end.
"""
# find next matfile
if not self._oneFullIter:
matfile = self._matfiles[np.random.randint(len(self._matfiles))]
else:
if self._fileidx >= len(self._matfiles):
self._fileidx = 0 # for next time
raise StopIteration
matfile = self._matfiles[self._fileidx]
self._fileidx += 1
# offset
offset = 0
if self._randoffset:
offset = np.random.randint(4)
# return features
if auto_bar == None:
return features.features_from_matfile(
matfile,
pSize=self._pSize,
usebars=self._usebars,
keyInv=self._keyInv,
songKeyInv=self._songKeyInv,
positive=self._positive,
do_resample=self._do_resample,
partialbar=self._partialbar,
offset=offset)
else:
# we assume auto_bar contains a model
# we predict on every offset until 4, return features with the best offset based on the model
best_dist = np.inf
best_feats = None
realSize = self._pSize
if self._partialbar > 0:
realSize = self._partialbar
# we go only until 4
for offset in range(min(realSize, 4)):
# get features
feats = features.features_from_matfile(
matfile,
pSize=self._pSize,
usebars=self._usebars,
keyInv=self._keyInv,
songKeyInv=self._songKeyInv,
positive=self._positive,
do_resample=self._do_resample,
partialbar=self._partialbar,
offset=offset)
if feats == None:
continue
# predicts
tmp, avg_dist = auto_bar.predicts(feats)
d = np.average(avg_dist)
if d < best_dist:
best_dist = d
best_feats = feats
# done, return best features
return best_feats
# analysis_dict2
analysis_dict2 = {'segstart':seg_start,'chromas':pitches,
'beatstart':beat_start,'barstart':bar_start,
'duration':duration}
tzan_feats = features.get_features(analysis_dict2,pSize=8,usebars=2,
keyInv=True,songKeyInv=False,
positive=False,do_resample=True,
btchroma_barbts=None)
tzan_feats = tzan_feats[np.nonzero(np.sum(tzan_feats,axis=1))]
print 'features from tzan data computed, shape =',tzan_feats.shape
# feature from matfile
mat_feats = features.features_from_matfile(tmpfilemat,pSize=8,usebars=2,
keyInv=True,songKeyInv=False,
positive=False,do_resample=True)
mat_feats = mat_feats[np.nonzero(np.sum(mat_feats,axis=1))]
print 'features from matfile computed, shape =',mat_feats.shape
# features from matfile old school
import data_iterator
import feats_utils as FU
data_iter = data_iterator.DataIterator()
data_iter.setMatfiles([tmpfilemat])
data_iter.useBars(2)
data_iter.stopAfterOnePass(True)
featsNorm = [FU.normalize_pattern_maxenergy(p,8,True,False).flatten() for p in data_iter]
featsNorm = np.array(featsNorm)
res = [np.sum(r) > 0 for r in featsNorm]
res2 = np.where(res)
def next_track(self, auto_bar=None):
"""
Returns features for a random matlab file, or None if some error.
If only one iter over the data, raise StopIteration at the end.
"""
# find next matfile
if not self._oneFullIter:
matfile = self._matfiles[np.random.randint(len(self._matfiles))]
else:
if self._fileidx >= len(self._matfiles):
self._fileidx = 0 # for next time
raise StopIteration
matfile = self._matfiles[self._fileidx]
self._fileidx += 1
# offset
offset = 0
if self._randoffset:
offset = np.random.randint(4)
# return features
if auto_bar == None:
return features.features_from_matfile(
matfile,
pSize=self._pSize,
usebars=self._usebars,
keyInv=self._keyInv,
songKeyInv=self._songKeyInv,
positive=self._positive,
do_resample=self._do_resample,
partialbar=self._partialbar,
offset=offset,
)
else:
# we assume auto_bar contains a model
# we predict on every offset until 4, return features with the best offset based on the model
best_dist = np.inf
best_feats = None
realSize = self._pSize
if self._partialbar > 0:
realSize = self._partialbar
# we go only until 4
for offset in range(min(realSize, 4)):
# get features
feats = features.features_from_matfile(
matfile,
pSize=self._pSize,
usebars=self._usebars,
keyInv=self._keyInv,
songKeyInv=self._songKeyInv,
positive=self._positive,
do_resample=self._do_resample,
partialbar=self._partialbar,
offset=offset,
)
if feats == None:
continue
# predicts
tmp, avg_dist = auto_bar.predicts(feats)
d = np.average(avg_dist)
if d < best_dist:
best_dist = d
best_feats = feats
# done, return best features
return best_feats
print 'do_resample =', do_resample
else:
break
sys.argv.pop(1)
# load song
songpath = sys.argv[1]
print 'song file:', songpath
dictpath = sys.argv[2]
print 'codebook file:', dictpath
# feats
feats = FEATS.features_from_matfile(songpath,
pSize=pSize,
usebars=usebars,
keyInv=keyInv,
songKeyInv=songKeyInv,
positive=positive,
do_resample=do_resample)
# model
mat = scipy.io.loadmat(dictpath)
codebook = mat['codebook']
model = MODEL.Model(codebook)
# predict
best_code_per_pattern, avg_dist = model.predicts(feats)
# report distortion (per... pixel? patch point?)
print 'average distortion:', np.average(avg_dist)
# build original and encoding
