resolved refactoring conflict

Author: ktaebum

README.md

@@ -44,3 +44,26 @@ Hyperparameter   | Value
 It seems that it can be trained more.  
 However, if train more, the reconstruction loss increase.  
 Based on this weird behavior, I recommand 15 epochs to train (max 20).
+
+## Dependency
+
+Refer [requirements](https://github.com/ktaebum/AttentionedDeepPaint/tree/master/requirements.txt)
+
+Install 
+1. Pytorch (>= 0.4.1)
+2. Torchvision (>= 0.2.1)
+based on your python version, os, cuda version etc...
+
+## Usage
+
+### Download dataset
+
+1. Create **data** folder
+2. go to [link](https://www.kaggle.com/ktaebum/animesketchcolorpair) and download
+3. unzip in **data** folder
+
+### Train
+
+`
+$ ./train.sh
+`

requirements.txt

@@ -0,0 +1,19 @@
+certifi==2018.10.15
+cloudpickle==0.6.1
+cycler==0.10.0
+dask==1.0.0
+decorator==4.3.0
+kiwisolver==1.0.1
+matplotlib==3.0.2
+networkx==2.2
+numpy==1.15.4
+Pillow==5.3.0
+pyparsing==2.3.0
+python-dateutil==2.7.5
+PyWavelets==1.0.1
+scikit-image==0.14.1
+scipy==1.1.0
+six==1.11.0
+toolz==0.9.0
+torch==0.4.1
+torchvision==0.2.1

train.sh

@@ -1,4 +1,2 @@
-python train.py --learning-rate 0.0002 --beta1 0.5 --verbose \
-       --batch-size 4 --save-every 5 --lambd 100 --model deepunet \
-       --sample 4 --no-mse --norm batch --num-epochs 20 --print-every 300 \
-       --train 
+python train.py --verbose --save-every 5 --model deepunet \
+       --sample 4 --no-mse --num-epochs 20 --print-every 300 \

trainer/deepunet.py

@@ -30,7 +30,11 @@ def __init__(self, *args):
         # log file
         if self.args.train:
             ctime = time.ctime().split()
+
             log_path = './log'
+            if not os.path.exists(log_path):
+                os.mkdir(log_path)
+
             log_dir = os.path.join(
                 log_path,
                 '%s_%s_%s_%s' % (ctime[-1], ctime[1], ctime[2], ctime[3]))
@@ -39,8 +43,11 @@ def __init__(self, *args):
                 f.write(str(args))
             self.log_file = open(os.path.join(log_dir, 'loss.txt'), 'w')
 
+        self.save_path = './data/result'
+        if not os.path.exists(self.save_path):
+            os.mkdir(self.save_path)
+
         # build model
-        self.resolution = self.args.resolution
         self.generator = DeepUNetPaintGenerator().to(self.device)
         self.discriminator = PatchGAN(sigmoid=self.args.no_mse).to(self.device)
 
@@ -215,19 +222,19 @@ def validate(self, dataset, epoch, samples=3):
                     (0, 0))
                 color_result.paste(
                     color2.crop((0, 0, self.resolution, self.resolution // 4)),
-                    (0, 512 // 4))
+                    (0, self.resolution // 4))
                 color_result.paste(
                     color3.crop((0, 0, self.resolution, self.resolution // 4)),
-                    (0, 512 // 4 * 2))
+                    (0, self.resolution // 4 * 2))
                 color_result.paste(
                     color4.crop((0, 0, self.resolution, self.resolution // 4)),
-                    (0, 512 // 4 * 3))
+                    (0, self.resolution // 4 * 3))
 
                 sub_result.paste(imageA, (0, 0))
-                sub_result.paste(styleB, (512, 0))
-                sub_result.paste(fakeB, (2 * 512, 0))
-                sub_result.paste(imageB, (3 * 512, 0))
-                sub_result.paste(color_result, (4 * 512, 0))
+                sub_result.paste(styleB, (self.resolution, 0))
+                sub_result.paste(fakeB, (2 * self.resolution, 0))
+                sub_result.paste(imageB, (3 * self.resolution, 0))
+                sub_result.paste(color_result, (4 * self.resolution, 0))
 
                 result.paste(sub_result, (0, 0 + self.resolution * i))
 
@@ -243,7 +250,7 @@ def validate(self, dataset, epoch, samples=3):
             save_image(
                 result,
                 'deepunetpaint_%03d_%02d' % (epoch, j),
-                './data/pair_niko/result',
+                self.save_path,
             )
 
     def test(self):

trainer/trainer.py

@@ -16,6 +16,7 @@ def __init__(self, args, data_loader, device):
         self.args = args
         self.data_loader = data_loader
         self.device = device
+        self.resolution = 512
 
     def train(self):
         raise NotImplementedError