Neural Feature

Extract feature from pre-trained deep model using Torch 7

Updates

• update function

Notes

Reading Videos in Torch 7

You can use ffmpeg package for reading videos in Torch 7, basically, you can first install the library by:

luarocks install ffmpeg

After the installation, you may find that you cannot read video because of lacking ffmpeg in your system (Torch 7's package and ffmepg build are different), Ubuntu 14.04 cannot find ffmepg because it removed in their official package maintain, you can install ffmpeg by following commands though:

sudo add-apt-repository ppa:mc3man/trusty-media
sudo apt-get update
sudo apt-get install ffmpeg

Now you can try this simple script to test your installation

require 'ffmpeg'

vid=ffmpeg.Video('/path/to/your/video');

vid:play{}

VGG ILSVRC-2014 16-layer

conv1_1: 64 3 3 3
conv1_2: 64 64 3 3
conv2_1: 128 64 3 3
conv2_2: 128 128 3 3
conv3_1: 256 128 3 3
conv3_2: 256 256 3 3
conv3_3: 256 256 3 3
conv4_1: 512 256 3 3
conv4_2: 512 512 3 3
conv4_3: 512 512 3 3
conv5_1: 512 512 3 3
conv5_2: 512 512 3 3
conv5_3: 512 512 3 3
fc6: 1 1 25088 4096
fc7: 1 1 4096 4096
fc8: 1 1 4096 1000
nn.Sequential {
  [input -> (1) -> (2) -> (3) -> (4) -> (5) -> (6) -> (7) -> (8) -> (9) -> (10) -> (11) -> (12) -> (13) -> (14) -> (15) -> (16) -> (17) -> (18) -> (19) -> (20) -> (21) -> (22) -> (23) -> (24) -> (25) -> (26) -> (27) -> (28) -> (29) -> (30) -> (31) -> (32) -> (33) -> (34) -> (35) -> (36) -> (37) -> (38) -> (39) -> (40) -> output]
  (1): nn.SpatialConvolutionMM
  (2): nn.ReLU
  (3): nn.SpatialConvolutionMM
  (4): nn.ReLU
  (5): inn.SpatialMaxPooling
  (6): nn.SpatialConvolutionMM
  (7): nn.ReLU
  (8): nn.SpatialConvolutionMM
  (9): nn.ReLU
  (10): inn.SpatialMaxPooling
  (11): nn.SpatialConvolutionMM
  (12): nn.ReLU
  (13): nn.SpatialConvolutionMM
  (14): nn.ReLU
  (15): nn.SpatialConvolutionMM
  (16): nn.ReLU
  (17): inn.SpatialMaxPooling
  (18): nn.SpatialConvolutionMM
  (19): nn.ReLU
  (20): nn.SpatialConvolutionMM
  (21): nn.ReLU
  (22): nn.SpatialConvolutionMM
  (23): nn.ReLU
  (24): inn.SpatialMaxPooling
  (25): nn.SpatialConvolutionMM
  (26): nn.ReLU
  (27): nn.SpatialConvolutionMM
  (28): nn.ReLU
  (29): nn.SpatialConvolutionMM
  (30): nn.ReLU
  (31): inn.SpatialMaxPooling
  (32): nn.View
  (33): nn.Linear(25088 -> 4096)
  (34): nn.ReLU
  (35): nn.Dropout
  (36): nn.Linear(4096 -> 4096)
  (37): nn.ReLU
  (38): nn.Dropout
  (39): nn.Linear(4096 -> 1000)
  (40): nn.SoftMax
}

Contacts

Yuhuang Hu
Advanced Robotic Lab
Department of Artificial Intelligence
Faculty of Computer Science & IT
University of Malaya
Email: duguyue100@gmail.com