User guide

Using as a feature extractor

In this example we take the output of the layer fc7 of the VGG face model as features.

>>> import numpy
>>> import bob.ip.caffe_extractor
>>> import bob.io.base
>>> from bob.io.base.test_utils import datafile
>>> image = bob.io.base.load(datafile('8821.hdf5', 'bob.ip.caffe_extractor'))
>>> caffe_extractor = bob.ip.caffe_extractor.VGGFace("fc7")
>>> numpy.allclose(
...     caffe_extractor(image)[0:5],
...     [ -0.55280668, 12.35865498, -1.54516554, -13.75179005, 2.49704885],
...     rtol=1e-04, atol=1e-05)
True

Note

The models will automatically download to the data folder of this package as soon as you start using them.

Using as a convolutional filter

In this example we plot some outputs of the convolutional layer conv2_1.

(Source code, png, hires.png, pdf)

_images/convolve.png