def get_labels(folder):
person_labels = []
io_labels = []
image_files = util.files(folder, '\d+.jpg$')
img_range = range(1, len(image_files) + 1)
for img_no in tqdm(img_range):
img = Image.open(folder + '/%04d.jpg' % (img_no))
io_labels.append(classify_indoors(img))
person_labels.append(classify_person(img))
img.close()
io_smoothed = gaussian(io_labels, sigma=6)
io_smoothed = [round(io_smooth) for io_smooth in io_smoothed]
person_smoothed = smooth(person_labels, 6)
return np.array(io_smoothed), np.array(person_smoothed)
def files_lastjob(self, extension = '*'):
"""Like files(), but only returns files that were created by the last job.
If no job has yet run on the document, returns the same files as the
files_uptodate() method.
"""
if self._start_time:
files = util.files(self.basenames(), extension)
try:
files = util.newer_files(files, self._start_time)
except (OSError, IOError):
pass
return files
else:
return self.files(extension)
def files(self, extension = '*', newer = True):
"""Returns a list of existing files matching our basenames and the given extension.
First the files basename + extension are returned,
then basename + '-[0-9]+' + extension,
then basename + '-.+' + extension.
If newer is True (the default), only files that are newer than the jobfile() are returned.
"""
jobfile = self.jobfile()
if jobfile:
files = util.files(self.basenames(), extension)
if newer:
try:
return util.newer_files(files, os.path.getmtime(jobfile))
except (OSError, IOError):
pass
return list(files)
return []
def _tryFetch(self, pkg, version):
mirror = pkg.distro.mirrorURL(pkg.dist, pkg.component)
pooldir = pkg.getCurrentSources()[0]['Directory']
name = "%s_%s.dsc" % (pkg.name, version)
url = "%s/%s/%s" % (mirror, pooldir, name)
ourPoolDir = pkg.poolDirectory()
outfile = "%s/%s" % (ourPoolDir.path, name)
logging.debug("Downloading %s to %s", url, outfile)
try:
self._getFile(url, outfile)
except IOError:
logging.debug("Could not download %s.", url)
return False
source = ControlFile()
try:
source.open(outfile, signed=True, multi_para=True)
except:
pass
for md5sum, size, name in files(source.paras[0]):
url = "%s/%s/%s" % (mirror, pooldir, name)
outfile = "%s/%s" % (ourPoolDir.path, name)
self._getFile(url, outfile, size)
ourPoolDir.updateSources()
return True
urls = np.load('dataset/urls.npy')
url_prefix = 'https://www.youtube.com/watch?v='
print('Num videos:', urls.shape[0])
failures = []
try:
mkdir('videos')
print('creating videos subdirectory')
except:
print('videos subdirectory already exists')
num_failed = 0
videos = util.files('dataset/videos')
sleep_time = 10
# remove extension to get ids
completed = [v[:-4]for v in videos]
remaining = set(urls) - set(completed)
print("Num Remaining: ",len(remaining))
for vid_id in remaining:
while True:
print('starting', vid_id)
url = url_prefix + vid_id
try:
print('url: ', url)
video = YouTube(url)
def test_files():
f = util.files('foo', '''
bar # comment
baz
''')
assert (f == ['foo/bar', 'foo/baz'])
models = files('', '''
blend
braninu # branin1 in GOMODELS
camel1u # camel1 in GOMODELS
chemeq
chi
gold # goldstein1 in GOMODELS
gridneta
griewank
hs105
hs106
hs109
hs111
hs112
hs114
hs116
hs15a
hs23
hs35
hs44
hs5
hs54
hs6
hs62
hs64
hs8
hs87
kowalik
levy3
ljcluster
osborne1
p2gon
pgon
powell
price
qb2
rosenbr
s324
s383
schwefel
shekel
steenbre
tre
weapon
''')
def test_files():
f = util.files('foo', '''
bar # comment
baz
''')
assert(f == ['foo/bar', 'foo/baz'])
if len(scores) == 0: scores = None
return (boxes, scores)
res = list(map(lam, class_indices))
arr = np.empty((5, 2), dtype='O')
arr[:] = res
return arr
episodes = [f.name for f in os.scandir(vid_location) if f.is_dir()]
results = {}
for ep in tqdm(episodes):
folder = f'{vid_location}/{ep}'
fils = util.files(folder, '(\d+).jpg')
inds = [int(re.match('(\d+).jpg', fil)[1]) for fil in fils]
full_files = [folder + '/' + f for f in fils]
res = list(zip(inds, full_files))
dataset = DetectorRealDataset(res, predictor=predictor)
loader = data.DataLoader(dataset, batch_size=4, num_workers=4)
generator = iter(loader)
ep_res = {}
for batch in generator:
ins = [{'image': im} for im in batch[1]]
res = predictor.model(ins)
vals = list(map(pred_to_score, res))
ind = batch[0]
for i in range(len(ind)):
ep_res[ind[i].item()] = vals[i]
results[ep] = ep_res
