def add_missing():
# For this testset, find how short of 10,000 we are
# Find that many new FpFiles that are not
# already part of our testset, add to testset (make sure >60sec)
files = db.session.query(db.FPFile).filter(db.FPFile.negative == False)\
.outerjoin(evaluation.Testfile).filter(evaluation.Testfile.id==None).all()
print "need %s candidate pos files. got %s to choose from" % (num_pos, len(files))
todo = []
random.shuffle(files)
while len(todo) < num_pos:
x = files.pop(0)
with audioread.audio_open(x.path.encode("utf-8")) as f:
duration = f.duration
print duration
if duration >= 60.0:
todo.append(x)
neg = db.session.query(db.FPFile).filter(db.FPFile.negative == True)\
.outerjoin(evaluation.Testfile).filter(evaluation.Testfile.id==None).all()
print "need %s candidate neg files. got %s to choose from" % (num_neg, len(neg))
random.shuffle(neg)
while len(todo) < num_pos+num_neg:
x = neg.pop(0)
with audioread.audio_open(x.path.encode("utf-8")) as f:
duration = f.duration
if duration >= 60.0:
todo.append(x)
print "adding %s files" % len(todo)
testset = db.session.query(evaluation.Testset).get(testset_id)
for fpfile in todo:
tfile = evaluation.Testfile(testset, fpfile)
db.session.add(tfile)
db.session.commit()
def short():
"""Actually delete bad testfiles (and all their results)"""
countneg = 0
countpos = 0
testset_id = 4
testfiles = db.session.query(evaluation.Testfile).filter(evaluation.Testfile.testset_id==testset_id)
print "Number testfiles: %s" % testfiles.count()
for i, tf in enumerate(testfiles):
if i % 100 == 0:
print i
with audioread.audio_open(tf.file.path.encode("utf-8")) as f:
duration = f.duration
if duration < 60.0:
if tf.file.negative:
countneg+=1
else:
countpos+=1
print "Removing short duration file: %s (%s)" % (tf.file.path.encode("utf-8"), duration)
cur = db.session.query(evaluation.Result).filter(evaluation.Result.testfile_id==tf.id)
print "%d results to remove" % cur.count()
cur.delete()
db.session.query(evaluation.Testfile).filter(evaluation.Testfile.id==tf.id).delete()
db.session.commit()
testfiles = db.session.query(evaluation.Testfile).filter(evaluation.Testfile.testset_id==testset_id)
print "New number testfiles: %s" % testfiles.count()
print "deleted negative: %s" % countneg
print "deleted positive: %s" % countpos
def getExecCommand(self, fromfile, tofile):
from chromaprint_support import audioread
try:
with audioread.audio_open(fromfile) as f:
if f.samplerate == 44100:
return ["sox", "-m", fromfile, self.mixfile, tofile, "trim", "0", "35"]
else:
c1 = ["sox", fromfile, "-t", "sox", "-", "trim", "0", "35", "rate", "44100"]
c2 = ["sox", "-m", "-t", "sox", "-", self.mixfile, tofile]
return (c1, c2)
except audioread.DecodeError:
return ["sox", "-m", fromfile, self.mixfile, tofile, "trim", "0", "35"]
def getExecCommand(self, fromfile, tofile):
from chromaprint_support import audioread
try:
with audioread.audio_open(fromfile) as f:
if f.samplerate == 44100:
return [
"sox", "-m", fromfile, self.mixfile, tofile, "trim",
"0", "35"
]
else:
c1 = [
"sox", fromfile, "-t", "sox", "-", "trim", "0", "35",
"rate", "44100"
]
c2 = ["sox", "-m", "-t", "sox", "-", self.mixfile, tofile]
return (c1, c2)
except audioread.DecodeError:
return [
"sox", "-m", fromfile, self.mixfile, tofile, "trim", "0", "35"
]
def rate():
""" Delete results for files that have a non-44.1k samplerate so we can re-do"""
testset_id = 4
c = 0
testfiles = db.session.query(evaluation.Testfile).filter(evaluation.Testfile.testset_id==testset_id)
print "Number testfiles: %s" % testfiles.count()
for i, tf in enumerate(testfiles):
if i % 100 == 0:
print i
with audioread.audio_open(tf.file.path.encode("utf-8")) as f:
rate = f.samplerate
if rate != 44100:
c += 1
print "Unexpected samplerate: %s (%s)" % (tf.file.path.encode("utf-8"), rate)
cur = db.session.query(evaluation.Result).filter(evaluation.Result.testfile_id==tf.id)
print "%d results to remove" % cur.count()
#cur.delete()
db.session.commit()
testfiles = db.session.query(evaluation.Testfile).filter(evaluation.Testfile.testset_id==testset_id)
print "to change", c
print "New number testfiles: %s" % testfiles.count()
def pre_lookup(self, file):
with audioread.audio_open(file) as f:
duration = f.duration
return {"duration": duration}
def pre_lookup(self, file):
with audioread.audio_open(file) as f:
duration = f.duration
return {"duration": duration}
