Merge pull request #8 from dionysio/master

Support for cmd line arguments, code refactoring

README.md

@@ -6,7 +6,7 @@ This algorithm is for
 * guided image synthesis (for example, transfer the style between different images)
 
 # Hardware
-* For cuda backend: choose 'speed' if your have at least 4GB graphic memory, and 'memory' otherwise. There is also an opencl backend (thanks to Dionýz Lazar). See "run_trans.lua" and "run_syn.lua" for our reference tests with Titan X, GT750M 2G and Sapphire Radeon R9 280 3G.  
+* For CUDA backend: choose 'speed' if your have at least 4GB graphic memory, and 'memory' otherwise. There is also an opencl backend (thanks to Dionýz Lazar). See "run_trans.lua" and "run_syn.lua" for our reference tests with Titan X, GT750M 2G and Sapphire Radeon R9 280 3G.
 
 
 # Examples
@@ -57,13 +57,13 @@ Pre-trained network:
 We use the the original VGG-19 model. You can find the download script at [Neural Style](https://github.com/jcjohnson/neural-style). The downloaded model and prototxt file MUST be saved in the folder "data/models"
 
 # Un-guided Synthesis
-* Run `qlua run_syn.lua` in a terminal. The algorithm will create a synthesis image of twice the size as the style input image.
+* Run `qlua cnnmrf.lua` in a terminal. Most important parameters are '-style_image' for specifying style input image and '-max_size' for resulting image size.
 * The content/style images are located in the folders "data/content" and "data/style" respectively. Notice by default the content image is the same as the style image; and the content image is only used for initalization (optional). 
 * Results are located in the folder "data/result/freesyn/MRF"
-* Parameters are defined & explained in "run_syn.lua".
+* All parameters are explained in "qlua cnnmrf.lua --help".
 
 # Guided Synthesis
-* Run `qlua run_trans.lua` in a terminal. The algorithm will synthesis using the texture of the style image and the structure of the content image. 
+* Run `qlua run_trans.lua` in a terminal. Most important parameters are '-style_image' for specifying style input image, '-content_image' for specifying content input image and '-max_size' for resulting image size.
 * The content/style images are located in the folders "data/content" and "data/style" respectively. 
 * Results are located in the folder "data/result/trans/MRF"
 * Parameters are defined & explained in "run_trans.lua".

cnnmrf.lua

@@ -0,0 +1,63 @@
+-- -*- coding: utf-8 -*-
+require 'torch'
+require 'paths'
+
+paths.dofile('mylib/helper.lua')
+
+-----------------------------------------
+-- Parameters
+-----------------------------------------
+
+
+
+cmd = torch.CmdLine()
+
+cmd:option('-content_name', 'potrait1', "The content image located in folder 'data/content'")
+cmd:option('-style_name', 'picasso', "The style image located in folder 'data/style'")
+cmd:option('-ini_method', 'image', "Initial method, set to 'image' to use the content image as the initialization; set to 'random' to use random noise.")
+cmd:option('-type', 'transfer', 'transfer|syn')
+cmd:option('-max_size',384, "Maximum size of the image. Larger image needs more time and memory.")
+cmd:option('-backend','cudnn', "Use cudnn' for CUDA-enabled GPUs or 'clnn' for OpenCL.")
+cmd:option('-mode','speed', "Try 'speed' if you have a GPU with more than 4GB memory, and try 'memory' otherwise. The 'speed' mode is significantly faster (especially for synthesizing high resolutions) at the cost of higher GPU memory. ")
+
+cmd:option('-num_res',3, "Number of resolutions. Notice the lowest resolution image should be larger than the patch size otherwise it won't synthesize.")
+cmd:option('-num_iter',{100, 100, 100}, "Number of iterations for each resolution.")
+cmd:option('-mrf_layers',{12, 21}, "The layers for MRF constraint. Usualy layer 21 alone already gives decent results. Including layer 12 may improve the results but at significantly more computational cost.")
+cmd:option('-mrf_weight',{1e-4, 1e-4}, "Weight for each MRF layer. Higher weights leads to more style faithful results.")
+cmd:option('-mrf_patch_size',{3, 3}, "The patch size for MRF constraint. This value is defined seperately for each MRF layer.")
+cmd:option('-target_num_rotation',0, 'To matching objects of different poses. This value is shared by all MRF layers. The total number of rotational copies is "2 * mrf_num_rotation + 1"')
+cmd:option('-target_num_scale',0, 'To matching objects of different scales. This value is shared by all MRF layers. The total number of scaled copies is "2 * mrf_num_scale + 1"')
+cmd:option('-target_sample_stride',{2, 2}, "Stride to sample mrf on style image. This value is defined seperately for each MRF layer.")
+cmd:option('-mrf_confidence_threshold',{0, 0}, "Threshold for filtering out bad matching. Default value 0 means we keep all matchings. This value is defined seperately for all layers.")
+cmd:option('-source_sample_stride',{2, 2}, "Stride to sample mrf on synthesis image. This value is defined seperately for each MRF layer. This settings is relevant only for syn setting.")
+
+cmd:option('-content_layers',{21}, "The layers for content constraint")
+cmd:option('-content_weight',2e1, "The weight for content constraint. Increasing this value will make the result more content faithful. Decreasing the value will make the method more style faithful. Notice this value should be increase (for example, doubled) if layer 12 is included for MRF constraint.")
+cmd:option('-tv_weight',1e-3, "TV smoothness weight")
+cmd:option('-scaler', 2, "Relative expansion from example to result. This settings is relevant only for syn setting.")
+
+cmd:option('-gpu_chunck_size_1',256, "Size of chunks to split feature maps along the channel dimension. This is to save memory when normalizing the matching score in mrf layers. Use large value if you have large gpu memory. As reference we use 256 for Titan X, and 32 for Geforce GT750M 2G.")
+cmd:option('-gpu_chunck_size_2',16, "Size of chuncks to split feature maps along the y dimension. This is to save memory when normalizing the matching score in mrf layers. Use large value if you have large gpu memory. As reference we use 16 for Titan X, and 2 for Geforce GT750M 2G.")
+
+-- fixed parameters
+cmd:option('-target_step_rotation', math.pi/24)
+cmd:option('-target_step_scale', 1.05)
+cmd:option('-output_folder', 'data/result/trans/MRF/')
+
+cmd:option('-proto_file', 'data/models/VGG_ILSVRC_19_layers_deploy.prototxt')
+cmd:option('-model_file', 'data/models/VGG_ILSVRC_19_layers.caffemodel')
+cmd:option('-gpu', 0, 'Zero-indexed ID of the GPU to use')
+cmd:option('-nCorrection', 25)
+cmd:option('-print_iter', 10)
+cmd:option('-save_iter', 10)
+
+params = cmd:parse(arg)
+
+local wrapper = nil
+if params.type == 'transfer' then
+    wrapper = require 'transfer_CNNMRF_wrapper'
+else
+    wrapper = require 'syn_CNNMRF_wrapper'
+end
+
+wrapper.main(params)

mylib/helper.lua

@@ -84,17 +84,17 @@ function computeBB(width, height, alpha)
 	local x0 = width / 2
 	local y0 = height / 2
 
-	x1r = x0+(x1-x0)*math.cos(alpha)+(y1-y0)*math.sin(alpha)
-	y1r = y0-(x1-x0)*math.sin(alpha)+(y1-y0)*math.cos(alpha)
+	local x1r = x0+(x1-x0)*math.cos(alpha)+(y1-y0)*math.sin(alpha)
+	local y1r = y0-(x1-x0)*math.sin(alpha)+(y1-y0)*math.cos(alpha)
 
-	x2r = x0+(x2-x0)*math.cos(alpha)+(y2-y0)*math.sin(alpha)
-	y2r = y0-(x2-x0)*math.sin(alpha)+(y2-y0)*math.cos(alpha)
+	local x2r = x0+(x2-x0)*math.cos(alpha)+(y2-y0)*math.sin(alpha)
+	local y2r = y0-(x2-x0)*math.sin(alpha)+(y2-y0)*math.cos(alpha)
 
-	x3r = x0+(x3-x0)*math.cos(alpha)+(y3-y0)*math.sin(alpha)
-	y3r = y0-(x3-x0)*math.sin(alpha)+(y3-y0)*math.cos(alpha)
+	local x3r = x0+(x3-x0)*math.cos(alpha)+(y3-y0)*math.sin(alpha)
+	local y3r = y0-(x3-x0)*math.sin(alpha)+(y3-y0)*math.cos(alpha)
 
-	x4r = x0+(x4-x0)*math.cos(alpha)+(y4-y0)*math.sin(alpha)
-	y4r = y0-(x4-x0)*math.sin(alpha)+(y4-y0)*math.cos(alpha)
+	local x4r = x0+(x4-x0)*math.cos(alpha)+(y4-y0)*math.sin(alpha)
+	local y4r = y0-(x4-x0)*math.sin(alpha)+(y4-y0)*math.cos(alpha)
 
 	-- print(x1r .. ' ' .. y1r .. ' ' .. x2r .. ' ' .. y2r .. ' ' .. x3r .. ' ' .. y3r .. ' ' .. x4r .. ' ' .. y4r)
 	if alpha > 0 then
@@ -139,15 +139,15 @@ function computeBB(width, height, alpha)
 end
 
 function computegrid(width, height, block_size, block_stride, flag_all)
-	coord_block_y = torch.range(1, height - block_size + 1, block_stride) 
+	local coord_block_y = torch.range(1, height - block_size + 1, block_stride)
 	if flag_all == 1 then
 		if coord_block_y[#coord_block_y] < height - block_size + 1 then
 		  local tail = torch.Tensor(1)
 		  tail[1] = height - block_size + 1
 		  coord_block_y = torch.cat(coord_block_y, tail)
 		end
 	end
-	coord_block_x = torch.range(1, width - block_size + 1, block_stride) 
+	local coord_block_x = torch.range(1, width - block_size + 1, block_stride)
 	if flag_all == 1 then
 		if coord_block_x[#coord_block_x] < width - block_size + 1 then
 		  local tail = torch.Tensor(1)
@@ -156,4 +156,30 @@ function computegrid(width, height, block_size, block_stride, flag_all)
 		end
 	end
 	return coord_block_x, coord_block_y
+end
+
+function preprocess(img)
+	local mean_pixel = torch.Tensor({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
+
+-- Undo the above preprocessing.
+function deprocess(img)
+	local mean_pixel = torch.Tensor({103.939, 116.779, 123.68})
+	mean_pixel = mean_pixel:view(3, 1, 1):expandAs(img)
+	img = img + mean_pixel:float()
+	local perm = torch.LongTensor{3, 2, 1}
+	img = img:index(1, perm):div(256.0)
+	return img
+end
+
+function run_tests(run_type, list_params)
+	local wrapper = run_type
+    for i_test = 1, #list_params do
+        wrapper.run_test(table.unpack(list_params[i_test]))
+    end
 end

mylib/mrf.lua

@@ -179,7 +179,7 @@ function MRFMM:updateGradInput(input, gradOutput)
   end
 
   -- local timer_AFT = torch.Timer()
-  max_response, max_id = torch.max(self.response, 1)
+  local max_response, max_id = torch.max(self.response, 1)
   -- local t_aft = timer_AFT:time().real
 
   -- local t_match = timer_MATCH:time().real

run_syn.lua

@@ -1,17 +1,8 @@
-require 'torch'
-require 'nn'
-require 'image'
 require 'paths'
-
-paths.dofile('mylib/myoptimizer.lua')
-paths.dofile('mylib/tv.lua')
-paths.dofile('mylib/mrf.lua')
 paths.dofile('mylib/helper.lua')
 
-syn_CNNMRF_wrapper = require 'syn_CNNMRF_wrapper'
-
 -----------------------------------------
--- parameters:
+-- Parameters:
 -----------------------------------------
 -- content_name: the content image located in folder "data/content". Notice for free synthesis this image is only used for initialization (and only when "ini_method" is set to "image")
 -- style_name: the style image located in folder "data/style" 
@@ -23,11 +14,11 @@ syn_CNNMRF_wrapper = require 'syn_CNNMRF_wrapper'
 
 -- mrf_layers: the layers for MRF constraint. Usualy layer 21 alone already gives decent results. Including layer 12 may improve the results but at significantly more computational cost.
 -- mrf_weight: weight for each MRF layer. Default value 1e-4. For free texture synthesis it can be seen as the "learning rate" in gradient decent.
--- mrf_patch_size: the patch size for MRF constraint. By default value 3. This value is defined seperately for each MRF layer.
+-- mrf_patch_size: the patch size for MRF constraint. Default value 3. This value is defined seperately for each MRF layer.
 -- mrf_num_rotation: To matching objects of different poses. Default value 0. This value is shared by all MRF layers. The total number of rotatoinal copies is "2 * mrf_num_rotation + 1"
 -- mrf_num_scale: To matching objects of different scales. Default value 0. This value is shared by all MRF layers. The total number of scaled copies is "2 * mrf_num_scale + 1"
--- mrf_sample_stride: stride to sample mrf on style image. Default value 2. This value is defined seperately for each MRF layer. This value is defined for each MRF layer.
--- mrf_synthesis_stride: stride to sample mrf on synthesis image. Default value 2. This value is defined seperately for each MRF layer. This value is defined for each MRF layer.
+-- mrf_sample_stride: stride to sample mrf on style image. Default value 2. This value is defined seperately for each MRF layer.
+-- mrf_synthesis_stride: stride to sample mrf on synthesis image. Default value 2. This value is defined seperately for each MRF layer.
 -- mrf_confidence_threshold: threshold for filtering out bad matching. Default value 0 -- means we keep all matchings. This value is defined seperately for all layers.
 
 -- tv_weight: TV smoothness weight. Default value 1e-3.
@@ -55,10 +46,4 @@ local list_params = {
                         {'2', '2', 'random', 384, 2, 3, {100, 100, 100}, {12, 21}, {1e-4, 1e-4}, {3, 3}, 1, 1, {2, 2}, {2, 2}, {0, 0}, 1e-3, 'memory', 256, 16, 'cudnn'},
                     }    
 
-for i_test = 1, #list_params do
-    local state = syn_CNNMRF_wrapper.state(list_params[i_test][1], list_params[i_test][2], list_params[i_test][3], list_params[i_test][4], list_params[i_test][5], list_params[i_test][6], list_params[i_test][7], list_params[i_test][8], list_params[i_test][9], list_params[i_test][10], list_params[i_test][11], list_params[i_test][12], list_params[i_test][13], list_params[i_test][14], list_params[i_test][15], list_params[i_test][16], list_params[i_test][17], list_params[i_test][18], list_params[i_test][19], list_params[i_test][20], list_params[i_test][21])
-    collectgarbage()
-end
-
-
-do return end
+run_tests(require 'syn_CNNMRF_wrapper', list_params)

run_trans.lua

@@ -1,18 +1,8 @@
-require 'torch'
-require 'nn'
-require 'image'
 require 'paths'
-
-paths.dofile('mylib/myoptimizer.lua')
-paths.dofile('mylib/tv.lua')
-paths.dofile('mylib/content.lua')
-paths.dofile('mylib/mrf.lua')
 paths.dofile('mylib/helper.lua')
 
-transfer_CNNMRF_wrapper = require 'transfer_CNNMRF_wrapper'
-
 -----------------------------------------
--- Parameters
+-- Parameters:
 -----------------------------------------
 -- content_name: the content image located in folder "data/content"
 -- style_name: the style image located in folder "data/style" 
@@ -77,12 +67,6 @@ local list_params = {
                     {'0', '0', 'image', 384, 3, {100, 100, 100}, {12, 21}, {1e-4, 1e-4}, {3, 3}, 3, 3, {2, 2}, {2, 2}, {0, 0}, {23}, 2e1, 1e-3, 'speed', 256, 16, 'cudnn'},
                     {'1', '1', 'image', 384, 3, {100, 100, 100}, {12, 21}, {1e-4, 1e-4}, {3, 3}, 3, 3, {2, 2}, {2, 2}, {0, 0}, {23}, 0.5e1, 1e-3, 'speed', 256, 16, 'cudnn'},
                     {'potrait1', 'picasso', 'image', 384, 3, {100, 100, 100}, {12, 21}, {1e-4, 1e-4}, {3, 3}, 1, 1, {2, 2}, {2, 2}, {0, 0}, {23}, 2e1, 1e-3, 'memory', 256, 16, 'clnn'},
-                    }
-
-for i_test = 1, #list_params do
-    local state = transfer_CNNMRF_wrapper.state(list_params[i_test][1], list_params[i_test][2], list_params[i_test][3], list_params[i_test][4], list_params[i_test][5], list_params[i_test][6], list_params[i_test][7], list_params[i_test][8], list_params[i_test][9], list_params[i_test][10], list_params[i_test][11], list_params[i_test][12], list_params[i_test][13], list_params[i_test][14], list_params[i_test][15], list_params[i_test][16], list_params[i_test][17], list_params[i_test][18], list_params[i_test][19], list_params[i_test][20], list_params[i_test][21])
-    collectgarbage()
-end
-
+}
 
-do return end
+run_tests(require 'transfer_CNNMRF_wrapper', list_params)