Module vipy.data.momentsintime

Classes

class MultiMoments (datadir)

Multi-Moments in Time: http://moments.csail.mit.edu/

>>> d = MultiMoments('/path/to/dir')
>>> valset = d.valset()
>>> valset.categories()           # return the dictionary mapping integer category to string
>>> valset[1].categories()        # return set of categories for this clip
>>> valset[1].category()          # return string encoded category for this clip (comma separated activity indexes)
>>> valset[1].play()              # Play the original clip 
>>> valset[1].mindim(224).show()  # Resize the clip to have minimum dimension 224, then show the modified clip
>>> valset[1].centersquare().mindim(112).saveas('out.mp4')  # modify the clip as square crop from the center with mindim=112, and save to new file
>>> valset[1].centersquare().mindim(112).normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)).torch(startframe=0, length=16)  # export 16x3x112x112 tensor

Expand source code Browse git

class MultiMoments(object):
    def __init__(self, datadir):
        """Multi-Moments in Time:  http://moments.csail.mit.edu/
        
          >>> d = MultiMoments('/path/to/dir')
          >>> valset = d.valset()
          >>> valset.categories()           # return the dictionary mapping integer category to string
          >>> valset[1].categories()        # return set of categories for this clip
          >>> valset[1].category()          # return string encoded category for this clip (comma separated activity indexes)
          >>> valset[1].play()              # Play the original clip 
          >>> valset[1].mindim(224).show()  # Resize the clip to have minimum dimension 224, then show the modified clip
          >>> valset[1].centersquare().mindim(112).saveas('out.mp4')  # modify the clip as square crop from the center with mindim=112, and save to new file
          >>> valset[1].centersquare().mindim(112).normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)).torch(startframe=0, length=16)  # export 16x3x112x112 tensor

        """
        self.datadir = datadir
        if not self._isdownloaded():
            raise ValueError('Not downloaded - datadir="/path/to/dir" such that "/path/to/dir/moments_categories.txt" exists.  See http://moments.csail.mit.edu/')
            
    def __repr__(self):
        return str('<vipy.data.multimoments: datadir="%s">' % (self.datadir))

    def _isdownloaded(self):
        return os.path.exists(os.path.join(self.datadir, 'moments_categories.txt'))
    
    def _dataset(self, csvfile):
        csv = readcsv(os.path.join(self.datadir, csvfile))
        categories = self.categories()        
        vidlist = []
        for r in csv:
            v = Scene(filename=os.path.join(self.datadir, 'videos', r[0]), category=','.join(sorted(r[1:])))
            for a in r[1:]:
                v.add(Activity(category=categories[int(a)], startframe=0, endframe=30*3))  # FIXME: framerate?
            vidlist.append(v)
        return vidlist

    def categories(self):
        return {int(r[1]):r[0] for r in readcsv(os.path.join(self.datadir, 'moments_categories.txt'))}

    def trainset(self):
        return self._dataset(os.path.join(self.datadir, 'trainingSet.txt'))  # FIXME: this is too slow due to ffmpeg-python hashing

    def valset(self):
        return self._dataset(os.path.join(self.datadir, 'validationSet.txt'))        

Methods

def categories(self)

Expand source code Browse git

def categories(self):
    return {int(r[1]):r[0] for r in readcsv(os.path.join(self.datadir, 'moments_categories.txt'))}

def trainset(self)

Expand source code Browse git

def trainset(self):
    return self._dataset(os.path.join(self.datadir, 'trainingSet.txt'))  # FIXME: this is too slow due to ffmpeg-python hashing

def valset(self)

Expand source code Browse git

def valset(self):
    return self._dataset(os.path.join(self.datadir, 'validationSet.txt'))