PN-Net: Conjoined Triple Deep Network for Learning Local Image Descriptors

Intro

Code for the arxiv paper PN-Net: Conjoined Triple Deep Network for Learning Local Image Descriptors. The network extracts feature descriptors from grayscale local patches of size 32x32.

Architecture

nn.Sequential {
  [input -> (1) -> (2) -> (3) -> (4) -> (5) -> (6) -> (7) -> (8) -> output]
  (1): cudnn.SpatialConvolution(1 -> 32, 7x7)
  (2): cudnn.Tanh
  (3): cudnn.SpatialMaxPooling(2,2,2,2)
  (4): cudnn.SpatialConvolution(32 -> 64, 6x6)
  (5): cudnn.Tanh
  (6): nn.View
  (7): nn.Linear(4096 -> 128)
  (8): cudnn.Tanh
}

Implementation details

For optimization details refer to the arxiv publication.

How to use

Download the sample dataset from http://vbalnt.io/notredame-torch.tar.gz and extract

Run th eval.lua

Efficiency

Efficiency comparison with MatchNet and deepcompare, both from CVPR 2015.