Spaces:
Running
Running
VDSR added
#6
by
namnguyen2103
- opened
- .gitattributes +1 -0
- app.py +241 -216
- models/VDSR/vdsr.ipynb +3 -0
- models/VDSR/vdsr.py +66 -0
- models/VDSR/vdsr_checkpoint.pth +3 -0
.gitattributes
CHANGED
@@ -37,3 +37,4 @@ models/HAT/hat-for-image-sr-2.ipynb filter=lfs diff=lfs merge=lfs -text
|
|
37 |
model/RCAN/rcan-for-image-sr.ipynb filter=lfs diff=lfs merge=lfs -text
|
38 |
models/RCAN/rcan-for-image-sr.ipynb filter=lfs diff=lfs merge=lfs -text
|
39 |
models/SRGAN/srgan-dl-prj.ipynb filter=lfs diff=lfs merge=lfs -text
|
|
|
|
37 |
model/RCAN/rcan-for-image-sr.ipynb filter=lfs diff=lfs merge=lfs -text
|
38 |
models/RCAN/rcan-for-image-sr.ipynb filter=lfs diff=lfs merge=lfs -text
|
39 |
models/SRGAN/srgan-dl-prj.ipynb filter=lfs diff=lfs merge=lfs -text
|
40 |
+
models/VDSR/vdsr.ipynb filter=lfs diff=lfs merge=lfs -text
|
app.py
CHANGED
@@ -1,216 +1,241 @@
|
|
1 |
-
import time
|
2 |
-
import streamlit as st
|
3 |
-
import subprocess
|
4 |
-
import numpy as np
|
5 |
-
from PIL import Image
|
6 |
-
from io import BytesIO
|
7 |
-
from models.HAT.hat import *
|
8 |
-
from models.RCAN.rcan import *
|
9 |
-
from models.SRGAN.srgan import *
|
10 |
-
from models.
|
11 |
-
from models.Interpolation.
|
12 |
-
from models.Interpolation.
|
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 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
st.
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
st.
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
st.session_state['
|
72 |
-
st.session_state['
|
73 |
-
st.session_state['
|
74 |
-
st.session_state['
|
75 |
-
st.session_state['
|
76 |
-
st.session_state['
|
77 |
-
st.session_state['
|
78 |
-
st.session_state['
|
79 |
-
st.session_state['
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
)
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
st.session_state['
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
st.session_state['
|
134 |
-
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
if st.
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
st.
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
|
191 |
-
st.
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
208 |
-
|
209 |
-
if st.
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import time
|
2 |
+
import streamlit as st
|
3 |
+
import subprocess
|
4 |
+
import numpy as np
|
5 |
+
from PIL import Image
|
6 |
+
from io import BytesIO
|
7 |
+
from models.HAT.hat import *
|
8 |
+
from models.RCAN.rcan import *
|
9 |
+
from models.SRGAN.srgan import *
|
10 |
+
from models.VDSR.vdsr import *
|
11 |
+
from models.Interpolation.nearest_neighbor import NearestNeighbor_for_deployment
|
12 |
+
from models.Interpolation.bilinear import Bilinear_for_deployment
|
13 |
+
from models.Interpolation.bicubic import Bicubic_for_deployment
|
14 |
+
|
15 |
+
subprocess.call('pip install natsort', shell=True)
|
16 |
+
from models.SRFlow.srflow import *
|
17 |
+
|
18 |
+
# Initialize session state for enhanced images
|
19 |
+
if 'nearest_enhanced_image' not in st.session_state:
|
20 |
+
st.session_state['nearest_enhanced_image'] = None
|
21 |
+
if 'bilinear_enhanced_image' not in st.session_state:
|
22 |
+
st.session_state['bilinear_enhanced_image'] = None
|
23 |
+
if 'bicubic_enhanced_image' not in st.session_state:
|
24 |
+
st.session_state['bicubic_enhanced_image'] = None
|
25 |
+
if 'hat_enhanced_image' not in st.session_state:
|
26 |
+
st.session_state['hat_enhanced_image'] = None
|
27 |
+
if 'rcan_enhanced_image' not in st.session_state:
|
28 |
+
st.session_state['rcan_enhanced_image'] = None
|
29 |
+
if 'srgan_enhanced_image' not in st.session_state:
|
30 |
+
st.session_state['srgan_enhanced_image'] = None
|
31 |
+
if 'srflow_enhanced_image' not in st.session_state:
|
32 |
+
st.session_state['srflow_enhanced_image'] = None
|
33 |
+
if 'vdsr_enhanced_image' not in st.session_state:
|
34 |
+
st.session_state['vdsr_enhanced_image'] = None
|
35 |
+
|
36 |
+
# Initialize session state for button clicks
|
37 |
+
if 'nearest_clicked' not in st.session_state:
|
38 |
+
st.session_state['nearest_clicked'] = False
|
39 |
+
if 'bilinear_clicked' not in st.session_state:
|
40 |
+
st.session_state['bilinear_clicked'] = False
|
41 |
+
if 'bicubic_clicked' not in st.session_state:
|
42 |
+
st.session_state['bicubic_clicked'] = False
|
43 |
+
if 'hat_clicked' not in st.session_state:
|
44 |
+
st.session_state['hat_clicked'] = False
|
45 |
+
if 'rcan_clicked' not in st.session_state:
|
46 |
+
st.session_state['rcan_clicked'] = False
|
47 |
+
if 'srgan_clicked' not in st.session_state:
|
48 |
+
st.session_state['srgan_clicked'] = False
|
49 |
+
if 'srflow_clicked' not in st.session_state:
|
50 |
+
st.session_state['srflow_clicked'] = False
|
51 |
+
if 'vdsr_clicked' not in st.session_state:
|
52 |
+
st.session_state['vdsr_clicked'] = False
|
53 |
+
|
54 |
+
st.markdown("<h1 style='text-align: center'>Image Super Resolution</h1>", unsafe_allow_html=True)
|
55 |
+
|
56 |
+
# Sidebar for navigation
|
57 |
+
st.sidebar.title("Options")
|
58 |
+
app_mode = st.sidebar.selectbox("Choose the input source", ["Upload image", "Take a photo"])
|
59 |
+
|
60 |
+
# Depending on the choice, show the uploader widget or webcam capture
|
61 |
+
if app_mode == "Upload image":
|
62 |
+
uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "png"], on_change=lambda: reset_states())
|
63 |
+
if uploaded_file is not None:
|
64 |
+
image = Image.open(uploaded_file).convert("RGB")
|
65 |
+
elif app_mode == "Take a photo":
|
66 |
+
camera_input = st.camera_input("Take a picture", on_change=lambda: reset_states())
|
67 |
+
if camera_input is not None:
|
68 |
+
image = Image.open(camera_input).convert("RGB")
|
69 |
+
|
70 |
+
def reset_states():
|
71 |
+
st.session_state['hat_enhanced_image'] = None
|
72 |
+
st.session_state['rcan_enhanced_image'] = None
|
73 |
+
st.session_state['srgan_enhanced_image'] = None
|
74 |
+
st.session_state['srflow_enhanced_image'] = None
|
75 |
+
st.session_state['bicubic_enhanced_image'] = None
|
76 |
+
st.session_state['bilinear_enhanced_image'] = None
|
77 |
+
st.session_state['nearest_enhanced_image'] = None
|
78 |
+
st.session_state['vdsr_enhanced_image'] = None
|
79 |
+
st.session_state['hat_clicked'] = False
|
80 |
+
st.session_state['rcan_clicked'] = False
|
81 |
+
st.session_state['srgan_clicked'] = False
|
82 |
+
st.session_state['srflow_clicked'] = False
|
83 |
+
st.session_state['bicubic_clicked'] = False
|
84 |
+
st.session_state['bilinear_clicked'] = False
|
85 |
+
st.session_state['nearest_clicked'] = False
|
86 |
+
st.session_state['vdsr_clicked'] = False
|
87 |
+
|
88 |
+
def get_image_download_link(img, filename):
|
89 |
+
"""Generates a link allowing the PIL image to be downloaded"""
|
90 |
+
# Convert the PIL image to Bytes
|
91 |
+
buffered = BytesIO()
|
92 |
+
img.save(buffered, format="PNG")
|
93 |
+
return st.download_button(
|
94 |
+
label="Download Image",
|
95 |
+
data=buffered.getvalue(),
|
96 |
+
file_name=filename,
|
97 |
+
mime="image/png"
|
98 |
+
)
|
99 |
+
|
100 |
+
if 'image' in locals():
|
101 |
+
# st.image(image, caption='Uploaded Image', use_column_width=True)
|
102 |
+
st.write("")
|
103 |
+
# ------------------------ Nearest Neighbor ------------------------ #
|
104 |
+
if st.button('Enhance with Nearest Neighbor'):
|
105 |
+
with st.spinner('Processing using Nearest Neighbor...'):
|
106 |
+
enhanced_image = NearestNeighbor_for_deployment(image)
|
107 |
+
st.session_state['nearest_enhanced_image'] = enhanced_image
|
108 |
+
st.session_state['nearest_clicked'] = True
|
109 |
+
st.success('Done!')
|
110 |
+
if st.session_state['nearest_enhanced_image'] is not None:
|
111 |
+
col1, col2 = st.columns(2)
|
112 |
+
col1.header("Original")
|
113 |
+
col1.image(image, use_column_width=True)
|
114 |
+
col2.header("Enhanced")
|
115 |
+
col2.image(st.session_state['nearest_enhanced_image'], use_column_width=True)
|
116 |
+
with col2:
|
117 |
+
get_image_download_link(st.session_state['nearest_enhanced_image'], 'nearest_enhanced.jpg')
|
118 |
+
|
119 |
+
# ------------------------ Bilinear ------------------------ #
|
120 |
+
if st.button('Enhance with Bilinear'):
|
121 |
+
with st.spinner('Processing using Bilinear...'):
|
122 |
+
enhanced_image = Bilinear_for_deployment(image)
|
123 |
+
st.session_state['bilinear_enhanced_image'] = enhanced_image
|
124 |
+
st.session_state['bilinear_clicked'] = True
|
125 |
+
st.success('Done!')
|
126 |
+
if st.session_state['bilinear_enhanced_image'] is not None:
|
127 |
+
col1, col2 = st.columns(2)
|
128 |
+
col1.header("Original")
|
129 |
+
col1.image(image, use_column_width=True)
|
130 |
+
col2.header("Enhanced")
|
131 |
+
col2.image(st.session_state['bilinear_enhanced_image'], use_column_width=True)
|
132 |
+
with col2:
|
133 |
+
get_image_download_link(st.session_state['bilinear_enhanced_image'], 'bilinear_enhanced.jpg')
|
134 |
+
|
135 |
+
# ------------------------ Bicubic ------------------------ #
|
136 |
+
if st.button('Enhance with Bicubic'):
|
137 |
+
with st.spinner('Processing using Bicubic...'):
|
138 |
+
enhanced_image = Bicubic_for_deployment(image)
|
139 |
+
st.session_state['bicubic_enhanced_image'] = enhanced_image
|
140 |
+
st.session_state['bicubic_clicked'] = True
|
141 |
+
st.success('Done!')
|
142 |
+
if st.session_state['bicubic_enhanced_image'] is not None:
|
143 |
+
col1, col2 = st.columns(2)
|
144 |
+
col1.header("Original")
|
145 |
+
col1.image(image, use_column_width=True)
|
146 |
+
col2.header("Enhanced")
|
147 |
+
col2.image(st.session_state['bicubic_enhanced_image'], use_column_width=True)
|
148 |
+
with col2:
|
149 |
+
get_image_download_link(st.session_state['bicubic_enhanced_image'], 'bicubic_enhanced.jpg')
|
150 |
+
|
151 |
+
# ------------------------ HAT ------------------------ #
|
152 |
+
if st.button('Enhance with HAT'):
|
153 |
+
with st.spinner('Processing using HAT...'):
|
154 |
+
with st.spinner('Wait for it... the model is processing the image'):
|
155 |
+
enhanced_image = HAT_for_deployment(image)
|
156 |
+
st.session_state['hat_enhanced_image'] = enhanced_image
|
157 |
+
st.session_state['hat_clicked'] = True
|
158 |
+
st.success('Done!')
|
159 |
+
if st.session_state['hat_enhanced_image'] is not None:
|
160 |
+
col1, col2 = st.columns(2)
|
161 |
+
col1.header("Original")
|
162 |
+
col1.image(image, use_column_width=True)
|
163 |
+
col2.header("Enhanced")
|
164 |
+
col2.image(st.session_state['hat_enhanced_image'], use_column_width=True)
|
165 |
+
with col2:
|
166 |
+
get_image_download_link(st.session_state['hat_enhanced_image'], 'hat_enhanced.jpg')
|
167 |
+
|
168 |
+
# ------------------------ RCAN ------------------------ #
|
169 |
+
if st.button('Enhance with RCAN'):
|
170 |
+
with st.spinner('Processing using RCAN...'):
|
171 |
+
with st.spinner('Wait for it... the model is processing the image'):
|
172 |
+
rcan_model = RCAN()
|
173 |
+
device = torch.device('cpu') if not torch.cuda.is_available() else torch.device('cuda')
|
174 |
+
rcan_model.load_state_dict(torch.load('models/RCAN/rcan_checkpoint.pth', map_location=device))
|
175 |
+
enhanced_image = rcan_model.inference(image)
|
176 |
+
st.session_state['rcan_enhanced_image'] = enhanced_image
|
177 |
+
st.session_state['rcan_clicked'] = True
|
178 |
+
st.success('Done!')
|
179 |
+
if st.session_state['rcan_enhanced_image'] is not None:
|
180 |
+
col1, col2 = st.columns(2)
|
181 |
+
col1.header("Original")
|
182 |
+
col1.image(image, use_column_width=True)
|
183 |
+
col2.header("Enhanced")
|
184 |
+
col2.image(st.session_state['rcan_enhanced_image'], use_column_width=True)
|
185 |
+
with col2:
|
186 |
+
get_image_download_link(st.session_state['rcan_enhanced_image'], 'rcan_enhanced.jpg')
|
187 |
+
|
188 |
+
# --------------------------SRGAN-------------------------- #
|
189 |
+
if st.button('Enhance with SRGAN'):
|
190 |
+
with st.spinner('Processing using SRGAN...'):
|
191 |
+
with st.spinner('Wait for it... the model is processing the image'):
|
192 |
+
srgan_model = GeneratorResnet()
|
193 |
+
device = torch.device('cpu') if not torch.cuda.is_available() else torch.device('cuda')
|
194 |
+
srgan_model = torch.load('models/SRGAN/srgan_checkpoint.pth', map_location=device)
|
195 |
+
enhanced_image = srgan_model.inference(image)
|
196 |
+
st.session_state['srgan_enhanced_image'] = enhanced_image
|
197 |
+
st.session_state['srgan_clicked'] = True
|
198 |
+
st.success('Done!')
|
199 |
+
if st.session_state['srgan_enhanced_image'] is not None:
|
200 |
+
col1, col2 = st.columns(2)
|
201 |
+
col1.header("Original")
|
202 |
+
col1.image(image, use_column_width=True)
|
203 |
+
col2.header("Enhanced")
|
204 |
+
col2.image(st.session_state['srgan_enhanced_image'], use_column_width=True)
|
205 |
+
with col2:
|
206 |
+
get_image_download_link(st.session_state['srgan_enhanced_image'], 'srgan_enhanced.jpg')
|
207 |
+
|
208 |
+
# ------------------------ SRFlow ------------------------ #
|
209 |
+
if st.button('Enhance with SRFlow'):
|
210 |
+
with st.spinner('Processing using SRFlow...'):
|
211 |
+
with st.spinner('Wait for it... the model is processing the image'):
|
212 |
+
enhanced_image = return_SRFlow_result(image)
|
213 |
+
st.session_state['srflow_enhanced_image'] = enhanced_image
|
214 |
+
st.session_state['srflow_clicked'] = True
|
215 |
+
st.success('Done!')
|
216 |
+
if st.session_state['srflow_enhanced_image'] is not None:
|
217 |
+
col1, col2 = st.columns(2)
|
218 |
+
col1.header("Original")
|
219 |
+
col1.image(image, use_column_width=True)
|
220 |
+
col2.header("Enhanced")
|
221 |
+
col2.image(st.session_state['srflow_enhanced_image'], use_column_width=True)
|
222 |
+
with col2:
|
223 |
+
get_image_download_link(st.session_state['srflow_enhanced_image'], 'srflow_enhanced.jpg')
|
224 |
+
|
225 |
+
# ------------------------ VDSR ------------------------ #
|
226 |
+
if st.button('Enhance with VDSR'):
|
227 |
+
with st.spinner('Processing using VDSR...'):
|
228 |
+
# Load the VDSR model
|
229 |
+
vdsr_model = torch.load('models/VDSR/vdsr_checkpoint.pth', map_location=torch.device('cpu'))
|
230 |
+
enhanced_image = vdsr_model.inference(image)
|
231 |
+
st.session_state['vdsr_enhanced_image'] = enhanced_image
|
232 |
+
st.session_state['vdsr_clicked'] = True
|
233 |
+
st.success('Done!')
|
234 |
+
if st.session_state['vdsr_enhanced_image'] is not None:
|
235 |
+
col1, col2 = st.columns(2)
|
236 |
+
col1.header("Original")
|
237 |
+
col1.image(image, use_column_width=True)
|
238 |
+
col2.header("Enhanced")
|
239 |
+
col2.image(st.session_state['vdsr_enhanced_image'], use_column_width=True)
|
240 |
+
with col2:
|
241 |
+
get_image_download_link(st.session_state['vdsr_enhanced_image'], 'vdsr_enhanced.jpg')
|
models/VDSR/vdsr.ipynb
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:acef3257ccc3ca0d7071b66a57f326b921e5696b917fe75c773afc08d14debcb
|
3 |
+
size 11173590
|
models/VDSR/vdsr.py
ADDED
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import torch.nn as nn
|
3 |
+
from torchvision.transforms import ToTensor
|
4 |
+
from PIL import Image
|
5 |
+
import os
|
6 |
+
from math import sqrt
|
7 |
+
import torch.nn.functional as F
|
8 |
+
|
9 |
+
#define class Block contain conv and relu layer
|
10 |
+
class Block(nn.Module):
|
11 |
+
def __init__(self):
|
12 |
+
super(Block, self).__init__()
|
13 |
+
self.conv = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, stride=1, padding=1, bias=False)
|
14 |
+
self.relu = nn.ReLU(inplace=True)
|
15 |
+
|
16 |
+
def forward(self, x):
|
17 |
+
return self.relu(self.conv(x))
|
18 |
+
|
19 |
+
class VDSR(nn.Module):
|
20 |
+
def __init__(self, in_channels=3, out_channels=3, num_blocks=18):
|
21 |
+
super(VDSR, self).__init__()
|
22 |
+
self.residual_layer = self.make_layer(Block, num_blocks)
|
23 |
+
self.input = nn.Conv2d(in_channels=in_channels, out_channels=64, kernel_size=3, stride=1, padding=1, bias=False)
|
24 |
+
self.output = nn.Conv2d(in_channels=64, out_channels=out_channels, kernel_size=3, stride=1, padding=1, bias=False)
|
25 |
+
self.relu = nn.ReLU(inplace=True)
|
26 |
+
|
27 |
+
for m in self.modules():
|
28 |
+
if isinstance(m, nn.Conv2d):
|
29 |
+
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
|
30 |
+
m.weight.data.normal_(0, sqrt(2. / n))
|
31 |
+
|
32 |
+
def make_layer(self, block, num_layers):
|
33 |
+
layers=[]
|
34 |
+
for _ in range(num_layers):
|
35 |
+
layers.append(block())
|
36 |
+
return nn.Sequential(*layers)
|
37 |
+
|
38 |
+
def forward(self, x):
|
39 |
+
residual = x
|
40 |
+
out = self.relu(self.input(x))
|
41 |
+
out = self.residual_layer(out)
|
42 |
+
out = self.output(out)
|
43 |
+
out = torch.add(residual, out)
|
44 |
+
return out
|
45 |
+
|
46 |
+
def inference(self, x):
|
47 |
+
"""
|
48 |
+
x is a PIL image
|
49 |
+
"""
|
50 |
+
self.eval()
|
51 |
+
with torch.no_grad():
|
52 |
+
x = ToTensor()(x).unsqueeze(0)
|
53 |
+
x = F.interpolate(x, scale_factor=4, mode='bicubic', align_corners=False).clamp(0, 1)
|
54 |
+
x = self.forward(x).clamp(0, 1)
|
55 |
+
x = Image.fromarray((x.squeeze(0).permute(1, 2, 0).detach().numpy() * 255).astype('uint8'))
|
56 |
+
return x
|
57 |
+
|
58 |
+
if __name__ == '__main__':
|
59 |
+
current_dir = os.path.dirname(os.path.realpath(__file__))
|
60 |
+
|
61 |
+
model = torch.load(current_dir + '/vdsr_checkpoint.pth', map_location=torch.device('cpu'))
|
62 |
+
model.eval()
|
63 |
+
with torch.no_grad():
|
64 |
+
input_image = Image.open('images/demo.png')
|
65 |
+
output_image = model.inference(input_image)
|
66 |
+
print(input_image.size, output_image.size)
|
models/VDSR/vdsr_checkpoint.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:839903f0373cbbd60ee00c4367436a718dd8689c8fda1d901471aa0f570e54be
|
3 |
+
size 2689946
|