def split(self):
trainPeople = [
'person02', 'person03', 'person05', 'person06', 'person07',
'person08', 'person09', 'person10', 'person22'
]
videos = self.dataset()
trainset = [
vid for vid in videos
if filetail(vid.filename()).split('_')[0] in trainPeople
]
testset = [
vid for vid in videos
if filetail(vid.filename()).split('_')[0] not in trainPeople
]
return (trainset, testset)
def imagetuplelist(list_of_tuples_of_image_files,
outdir,
title='Image Visualization',
imagewidth=64):
"""Imageset but put tuples on same row"""
k_divid = 0
# Create summary page to show precomputed images
outdir = remkdir(outdir)
filename = os.path.join(remkdir(outdir), 'index.html')
f = open(filename, 'w')
f.write('<!DOCTYPE html>\n')
f.write('<html>\n')
f.write('<body>\n')
f.write('<div id="container" style="width:2400px">\n')
f.write('<div id="header">\n')
f.write('<h1 style="margin-bottom:0;">Title: %s</h1><br>\n' % title)
localtime = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(time.time()))
f.write('Summary HTML generated on %s<br>\n' % localtime)
f.write('Number of Tuples: %d<br>\n' % len(list_of_tuples_of_image_files))
f.write('</div>\n')
f.write('<br>\n')
f.write('<hr>\n')
f.write('<div id="%04d" style="float:left;">\n' % k_divid)
k_divid = k_divid + 1
# Generate images and html
for (k, imsrclist) in enumerate(list_of_tuples_of_image_files):
f.write('<p>\n</p>\n')
for imsrc in imsrclist:
shutil.copyfile(imsrc, os.path.join(outdir, filetail(imsrc)))
imdst = filetail(imsrc)
f.write('<b>Filename: %s</b><br>\n' % imdst)
f.write('<p>\n</p>\n')
f.write('<br>\n')
for imsrc in imsrclist:
imdst = filetail(imsrc)
f.write('<img src="%s" alt="image" width=%d/>' %
(imdst, imagewidth))
f.write('\n<p>\n</p>\n')
f.write('<hr>\n')
f.write('<p>\n</p>\n')
f.write('</div>\n')
f.write('</body>\n')
f.write('</html>\n')
f.close()
return filename
def testset(self):
"""ActivityNet test set does not include any annotations"""
assert self._isdownloaded(
), "Dataset not downloaded. download() first or manually download '%s' into '%s'" % (
self._url, self.datadir)
json = readjson(os.path.join(self.datadir, filetail(URL)))
return [
Video(url=v['url'], filename=os.path.join(self.datadir, youtubeid))
for (youtubeid, v) in json['database'].items()
if v['subset'] == 'testing'
]
def videos(self, subject):
videos = {}
for d in dirlist(os.path.join(self.ytfdir, 'frame_images_DB',
subject)):
k_videoindex = filetail(d)
videos[k_videoindex] = []
for f in imlist(d):
videos[k_videoindex].append(
ImageCategory(filename=f, category=subject))
videos[k_videoindex] = sorted(videos[k_videoindex],
key=lambda im: im.filename())
return videos
def _dataset(self, subset):
assert self._isdownloaded(
), "Dataset not downloaded. download() first or manually download '%s' into '%s'" % (
self._url, self.datadir)
jsonfile = os.path.join(self.datadir, filetail(URL))
json = readjson(jsonfile)
return [
VideoCategory(url=v['url'],
filename=os.path.join(self.datadir, youtubeid),
category=a['label'],
startsec=float(a['segment'][0]),
endsec=float(a['segment'][1]))
for (youtubeid, v) in json['database'].items()
for a in v['annotations'] if v['subset'] == subset
]
def fetchjson(self):
"""Download JSON if not already downloaded"""
if self._jsonfile is None:
self._jsonfile = os.path.join(
remkdir(os.environ["VIPY_CACHE"] if "VIPY_CACHE" in
os.environ else tempdir()),
filetail(self._jsonurl),
)
if not os.path.exists(self._jsonfile):
print('[pycollector.video]: Fetching "%s"' % self._jsonurl)
try:
vipy.downloader.s3(
self._jsonurl, self._jsonfile
) # TODO - this is a problem to assume vipy user also has access to S3. We should decouple this dependency of using vipy
except KeyboardInterrupt:
raise
except Exception as e:
print(
'[pycollector.video]: S3 download error "%s" - SKIPPING'
% str(e))
jsonfile = None
return self
def download(self):
vipy.downloader.download(URL, os.path.join(self.datadir,
filetail(URL)))
self._unpack(os.path.join(self.datadir, filetail(URL)), self.datadir)
def subjects(self):
"""List of all subject names"""
return [filetail(d) for d in dirlist(self.lfwdir)]
def download(self):
vipy.downloader.download(URL, os.path.join(self.datadir,
filetail(URL)))
return self
def download(self):
zipfile = os.path.join(self.datadir, filetail(URL))
vipy.downloader.download(URL, zipfile)
vipy.downloader.unpack(zipfile, self.datadir)
return self