bestchannels.lua

-- this program takes in an image
-- and finds nc channels on a given layer
-- having the strongest activations 

require 'torch'
require 'nn'
require 'image'
require 'loadcaffe' 

function preprocess(img)
   local mean_pixel = torch.DoubleTensor({103.939, 116.779, 123.68})
   local perm = torch.LongTensor{3, 2, 1}
   img = img:index(1, perm):mul(256.0)
   mean_pixel = mean_pixel:view(3, 1, 1):expandAs(img)
   img:add(-1, mean_pixel)
   return img
end

local cmd = torch.CmdLine()

cmd:option('-image', 'examples/inputs/tubingen.jpg')
cmd:option('-image_size', 800, 'output image size')
cmd:option('-proto', 'models/VGG_ILSVRC_19_layers_deploy.prototxt')
cmd:option('-model', 'models/VGG_ILSVRC_19_layers.caffemodel')
cmd:option('-layer', 'relu4_2', 'layer for examine')
cmd:option('-nc', '10', 'number of best channels to be shown')

local params = cmd:parse(arg)

local content_image = image.load(params.image, 3)
content_image = image.scale(content_image, params.image_size, 'bilinear')
local content_image_caffe = preprocess(content_image):float()
local img = content_image_caffe:clone():float()


local cnn = loadcaffe.load(params.proto, params.model, "nn"):float()

local net = nn.Sequential()


for i = 1, #cnn do
      local layer = cnn:get(i)
      local typ = torch.type(layer)
      local name = layer.name
      print(name, typ)
      net:add(layer)
      if (name == params.layer) then break end
      if (i == #cnn) then 
        print("No such layer: "..params.layer)
        return 
      end   
end

local y = net:forward(img)

local n = y:size(1)
activ = torch.Tensor(n)

for i = 1, n do
  local y3 = torch.Tensor(3,y:size(2),y:size(3))
  local y1 = y:clone():narrow(1,i,1)
  local norm = torch.norm(y1)
  activ[i] = norm   
end 

channels, idx = torch.sort(activ, 1, true)
for i = 1, params.nc do
  print(idx[i], channels[i])
end