Commit @97e1e42f0ac512996e82807a606d568732ee375c

윤영준 2023-07-24

typo

@97e1e42f0ac512996e82807a606d568732ee375c

fdc2bde

97e1e42

model/AttentiveRNN.py

--- model/AttentiveRNN.py

+++ model/AttentiveRNN.py


         attention_map = torch.sigmoid(attention_map)
 
         ret = {
-            "attention_amp" : attention_map,
+            "attention_map" : attention_map,
             "cell_state" : cell_state,
             "lstm_feats" : lstm_feats
         }

fdc2bde

97e1e42

train.py

--- train.py

+++ train.py


 import subprocess
 import atexit
 import torchvision.transforms
+import cv2
+
 from visdom import Visdom
 from torchvision.utils import save_image
+from torchvision import transforms
+from torchvision.transforms import RandomCrop, RandomPerspective, Compose
 from torch.utils.data import DataLoader
 from time import gmtime, strftime
 

 from tools.argparser import get_param
 from tools.logger import Logger
 from tools.dataloader import ImagePairDataset
-
-
-# this function is from https://github.com/eriklindernoren/PyTorch-GAN/blob/master/implementations/dualgan/models.py
-# MIT license
-def weights_init_normal(m):
-    classname = m.__class__.__name__
-    if classname.find("Conv") != -1:
-        torch.nn.init.normal_(m.weight.data, 0.0, 0.02)
-    elif classname.find("BatchNorm2d") != -1:
-        torch.nn.init.normal_(m.weight.data, 1.0, 0.02)
-        torch.nn.init.constant_(m.bias.data, 0.0)
 
 args = get_param()
 # I am doing this for easier debugging

 clean_data_path = glob.glob("data/source/Oxford_raindrop_dataset/clean/*.png")
 clean_data_path = sorted(clean_data_path)
 
-resize = torchvision.transforms.Resize((480, 720), antialias=True)
-dataset = ImagePairDataset(clean_data_path, rainy_data_path, transform=resize)
+height = 480
+width  = 720
+
+transform = Compose([
+    RandomPerspective(),
+    RandomCrop((height, width))
+])
+
+resize = torchvision.transforms.Resize((height, width), antialias=True)
+dataset = ImagePairDataset(clean_data_path, rainy_data_path, transform=transform)
 dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True, num_workers=4)
 # declare generator loss
 
 optimizer_G = torch.optim.Adam(generator.parameters(), lr=generator_learning_rate)
-optimizer_G_ARNN = torch.optim.Adam(generator.attentiveRNN.parameters(), lr=generator_ARNN_learning_rate)
-optimizer_G_AE = torch.optim.Adam(generator.autoencoder.parameters(), lr=generator_learning_rate)
+optimizer_G_ARNN = torch.optim.Adam(generator.attentiveRNN.parameters(), lr=generator_ARNN_learning_rate, betas=(0.5, 0.999))
+optimizer_G_AE = torch.optim.Adam(generator.autoencoder.parameters(), lr=generator_learning_rate, betas=(0.5, 0.999))
 optimizer_D = torch.optim.Adam(discriminator.parameters(), lr=discriminator_learning_rate)
 
 # ------visdom visualizer ----------

 ARNN_loss_window = vis.line(Y=np.array([0]), X=np.array([0]), opts=dict(title='Generator-AttentionRNN Loss'))
 AE_loss_window = vis.line(Y=np.array([0]), X=np.array([0]), opts=dict(title='Generator-AutoEncoder Loss'))
 Discriminator_loss_window = vis.line(Y=np.array([0]), X=np.array([0]), opts=dict(title='Discriminator Loss'))
-Attention_map_visualizer = vis.image(np.zeros((692, 776)), opts=dict(title='Attention Map'))
-Difference_mask_map_visualizer = vis.image(np.zeros((692,776)), opts=dict(title='Mask Map'))
-Generator_output_visualizer = vis.image(np.zeros((692,776)), opts=dict(title='Generated Derain Output'))
-Input_image_visualizer = vis.image(np.zeros((692,776)), opts=dict(title='input clean image'))
+Attention_map_visualizer = vis.image(np.zeros((height, width)), opts=dict(title='Attention Map'))
+Difference_mask_map_visualizer = vis.image(np.zeros((height, width)), opts=dict(title='Mask Map'))
+Generator_output_visualizer = vis.image(np.zeros((height, width)), opts=dict(title='Generated Derain Output'))
+Input_image_visualizer = vis.image(np.zeros((height, width)), opts=dict(title='input clean image'))
 
 for epoch_num, epoch in enumerate(range(epochs)):
     for i, imgs in enumerate(dataloader):

 
         logger.print_training_log(epoch_num, epochs, i, len(dataloader), losses)
         # visdom logger
-        vis.line(Y=np.array([generator_loss_ARNN.item()]), X=np.array([epoch * epoch_num + i]), win=ARNN_loss_window,
+        vis.line(Y=np.array([generator_loss_ARNN.item()]), X=np.array([epoch_num * epochs + i]), win=ARNN_loss_window,
                  update='append')
-        vis.line(Y=np.array([generator_loss_AE.item()]), X=np.array([epoch * epoch_num + i]), win=AE_loss_window,
+        vis.line(Y=np.array([generator_loss_AE.item()]), X=np.array([epoch_num * epochs + i]), win=AE_loss_window,
                  update='append')
-        vis.line(Y=np.array([discriminator_loss.item()]), X=np.array([epoch * epoch_num + i]), win=Discriminator_loss_window,
+        vis.line(Y=np.array([discriminator_loss.item()]), X=np.array([epoch_num * epochs + i]), win=Discriminator_loss_window,
                  update='append')
         vis.image(generator_attention_map[-1][0, 0, :, :], win=Attention_map_visualizer,
                   opts=dict(title="Attention Map"))

Add a comment

Open 0
Closed 0

List

...	...	@@ -101,7 +101,7 @@
101	101	attention_map = torch.sigmoid(attention_map)
102	102
103	103	ret = {
104		- "attention_amp" : attention_map,
	104	+ "attention_map" : attention_map,
105	105	"cell_state" : cell_state,
106	106	"lstm_feats" : lstm_feats
107	107	}

...	...	@@ -8,8 +8,12 @@
8	8	import subprocess
9	9	import atexit
10	10	import torchvision.transforms
	11	+import cv2
	12	+
11	13	from visdom import Visdom
12	14	from torchvision.utils import save_image
	15	+from torchvision import transforms
	16	+from torchvision.transforms import RandomCrop, RandomPerspective, Compose
13	17	from torch.utils.data import DataLoader
14	18	from time import gmtime, strftime
15	19
...	...	@@ -20,17 +24,6 @@
20	24	from tools.argparser import get_param
21	25	from tools.logger import Logger
22	26	from tools.dataloader import ImagePairDataset
23		-
24		-
25		-# this function is from https://github.com/eriklindernoren/PyTorch-GAN/blob/master/implementations/dualgan/models.py
26		-# MIT license
27		-def weights_init_normal(m):
28		- classname = m.__class__.__name__
29		- if classname.find("Conv") != -1:
30		- torch.nn.init.normal_(m.weight.data, 0.0, 0.02)
31		- elif classname.find("BatchNorm2d") != -1:
32		- torch.nn.init.normal_(m.weight.data, 1.0, 0.02)
33		- torch.nn.init.constant_(m.bias.data, 0.0)
34	27
35	28	args = get_param()
36	29	# I am doing this for easier debugging
...	...	@@ -71,14 +64,22 @@
71	64	clean_data_path = glob.glob("data/source/Oxford_raindrop_dataset/clean/*.png")
72	65	clean_data_path = sorted(clean_data_path)
73	66
74		-resize = torchvision.transforms.Resize((480, 720), antialias=True)
75		-dataset = ImagePairDataset(clean_data_path, rainy_data_path, transform=resize)
	67	+height = 480
	68	+width = 720
	69	+
	70	+transform = Compose([
	71	+ RandomPerspective(),
	72	+ RandomCrop((height, width))
	73	+])
	74	+
	75	+resize = torchvision.transforms.Resize((height, width), antialias=True)
	76	+dataset = ImagePairDataset(clean_data_path, rainy_data_path, transform=transform)
76	77	dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True, num_workers=4)
77	78	# declare generator loss
78	79
79	80	optimizer_G = torch.optim.Adam(generator.parameters(), lr=generator_learning_rate)
80		-optimizer_G_ARNN = torch.optim.Adam(generator.attentiveRNN.parameters(), lr=generator_ARNN_learning_rate)
81		-optimizer_G_AE = torch.optim.Adam(generator.autoencoder.parameters(), lr=generator_learning_rate)
	81	+optimizer_G_ARNN = torch.optim.Adam(generator.attentiveRNN.parameters(), lr=generator_ARNN_learning_rate, betas=(0.5, 0.999))
	82	+optimizer_G_AE = torch.optim.Adam(generator.autoencoder.parameters(), lr=generator_learning_rate, betas=(0.5, 0.999))
82	83	optimizer_D = torch.optim.Adam(discriminator.parameters(), lr=discriminator_learning_rate)
83	84
84	85	# ------visdom visualizer ----------
...	...	@@ -93,10 +94,10 @@
93	94	ARNN_loss_window = vis.line(Y=np.array([0]), X=np.array([0]), opts=dict(title='Generator-AttentionRNN Loss'))
94	95	AE_loss_window = vis.line(Y=np.array([0]), X=np.array([0]), opts=dict(title='Generator-AutoEncoder Loss'))
95	96	Discriminator_loss_window = vis.line(Y=np.array([0]), X=np.array([0]), opts=dict(title='Discriminator Loss'))
96		-Attention_map_visualizer = vis.image(np.zeros((692, 776)), opts=dict(title='Attention Map'))
97		-Difference_mask_map_visualizer = vis.image(np.zeros((692,776)), opts=dict(title='Mask Map'))
98		-Generator_output_visualizer = vis.image(np.zeros((692,776)), opts=dict(title='Generated Derain Output'))
99		-Input_image_visualizer = vis.image(np.zeros((692,776)), opts=dict(title='input clean image'))
	97	+Attention_map_visualizer = vis.image(np.zeros((height, width)), opts=dict(title='Attention Map'))
	98	+Difference_mask_map_visualizer = vis.image(np.zeros((height, width)), opts=dict(title='Mask Map'))
	99	+Generator_output_visualizer = vis.image(np.zeros((height, width)), opts=dict(title='Generated Derain Output'))
	100	+Input_image_visualizer = vis.image(np.zeros((height, width)), opts=dict(title='input clean image'))
100	101
101	102	for epoch_num, epoch in enumerate(range(epochs)):
102	103	for i, imgs in enumerate(dataloader):
...	...	@@ -153,11 +154,11 @@
153	154
154	155	logger.print_training_log(epoch_num, epochs, i, len(dataloader), losses)
155	156	# visdom logger
156		- vis.line(Y=np.array([generator_loss_ARNN.item()]), X=np.array([epoch * epoch_num + i]), win=ARNN_loss_window,
	157	+ vis.line(Y=np.array([generator_loss_ARNN.item()]), X=np.array([epoch_num * epochs + i]), win=ARNN_loss_window,
157	158	update='append')
158		- vis.line(Y=np.array([generator_loss_AE.item()]), X=np.array([epoch * epoch_num + i]), win=AE_loss_window,
	159	+ vis.line(Y=np.array([generator_loss_AE.item()]), X=np.array([epoch_num * epochs + i]), win=AE_loss_window,
159	160	update='append')
160		- vis.line(Y=np.array([discriminator_loss.item()]), X=np.array([epoch * epoch_num + i]), win=Discriminator_loss_window,
	161	+ vis.line(Y=np.array([discriminator_loss.item()]), X=np.array([epoch_num * epochs + i]), win=Discriminator_loss_window,
161	162	update='append')
162	163	vis.image(generator_attention_map[-1][0, 0, :, :], win=Attention_map_visualizer,
163	164	opts=dict(title="Attention Map"))

Delete comment