Example: barber
Tech report (v5) - arXiv

Tech report (v5) - arXiv

Back to document page

Feature extraction. We extract a 4096-dimensional fea-ture vector from each region proposal using the Caffe [24] implementation of the CNN described by Krizhevsky et al. [25]. Features are computed by forward propagating a mean-subtracted 227 227 RGB image through five con-volutional layers and two fully connected layers. We refer

  Feature, Using, True, Fea ture

Download Tech report (v5) - arXiv


Information

Domain:

Source:

Link to this page:

Please notify us if you found a problem with this document:

Other abuse

Advertisement

Related search queries