Update app.py
Browse files
app.py
CHANGED
|
@@ -6,16 +6,20 @@ import numpy as np
|
|
| 6 |
from PIL import Image
|
| 7 |
from model.flol import create_model
|
| 8 |
|
| 9 |
-
|
| 10 |
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
|
| 11 |
#define some auxiliary functions
|
| 12 |
pil_to_tensor = transforms.ToTensor()
|
| 13 |
|
| 14 |
-
# Define a
|
| 15 |
-
|
| 16 |
-
"
|
| 17 |
-
"
|
| 18 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
# Initial model setup (without weights)
|
| 21 |
model = create_model()
|
|
@@ -25,10 +29,10 @@ def load_img(filename):
|
|
| 25 |
img_tensor = pil_to_tensor(img)
|
| 26 |
return img_tensor
|
| 27 |
|
| 28 |
-
def process_img(image,
|
| 29 |
-
#
|
| 30 |
-
if
|
| 31 |
-
model_path =
|
| 32 |
checkpoints = torch.load(model_path, map_location=device)
|
| 33 |
model.load_state_dict(checkpoints['params'])
|
| 34 |
model.to(device)
|
|
@@ -59,9 +63,9 @@ Due to the GPU memory limitations, the app might crash if you feed a high-resolu
|
|
| 59 |
'''
|
| 60 |
|
| 61 |
examples = [['images/425_UHD_LL.JPG'],
|
| 62 |
-
['images/
|
| 63 |
-
['images/
|
| 64 |
-
['images/
|
| 65 |
['images/1778_UHD_LL.JPG'],
|
| 66 |
['images/1791_UHD_LL.JPG']]
|
| 67 |
|
|
@@ -76,8 +80,8 @@ css = """
|
|
| 76 |
demo = gr.Interface(
|
| 77 |
fn=process_img,
|
| 78 |
inputs=[
|
| 79 |
-
gr.Image(type='pil', label='input'),
|
| 80 |
-
gr.
|
| 81 |
],
|
| 82 |
outputs=[gr.Image(type='pil', label='output')],
|
| 83 |
title=title,
|
|
@@ -86,5 +90,16 @@ demo = gr.Interface(
|
|
| 86 |
css=css
|
| 87 |
)
|
| 88 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 89 |
if __name__ == '__main__':
|
| 90 |
demo.launch()
|
|
|
|
| 6 |
from PIL import Image
|
| 7 |
from model.flol import create_model
|
| 8 |
|
|
|
|
| 9 |
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
|
| 10 |
#define some auxiliary functions
|
| 11 |
pil_to_tensor = transforms.ToTensor()
|
| 12 |
|
| 13 |
+
# Define a dictionary to map image filenames to weight files
|
| 14 |
+
image_to_weights = {
|
| 15 |
+
"425_UHD_LL.JPG": './weights/flolv2_UHDLL.pt',
|
| 16 |
+
"1778_UHD_LL.JPG": './weights/flolv2_UHDLL.pt',
|
| 17 |
+
"1791_UHD_LL.JPG": './weights/flolv2_UHDLL.pt',
|
| 18 |
+
|
| 19 |
+
"low00748.png": './weights/flolv2_all_111439.pt',
|
| 20 |
+
"low00723.png": './weights/flolv2_all_111439.pt',
|
| 21 |
+
"low00772.png": './weights/flolv2_all_111439.pt'
|
| 22 |
+
}
|
| 23 |
|
| 24 |
# Initial model setup (without weights)
|
| 25 |
model = create_model()
|
|
|
|
| 29 |
img_tensor = pil_to_tensor(img)
|
| 30 |
return img_tensor
|
| 31 |
|
| 32 |
+
def process_img(image, filename):
|
| 33 |
+
# Select the correct weight file based on the image filename
|
| 34 |
+
if filename in image_to_weights:
|
| 35 |
+
model_path = image_to_weights[filename]
|
| 36 |
checkpoints = torch.load(model_path, map_location=device)
|
| 37 |
model.load_state_dict(checkpoints['params'])
|
| 38 |
model.to(device)
|
|
|
|
| 63 |
'''
|
| 64 |
|
| 65 |
examples = [['images/425_UHD_LL.JPG'],
|
| 66 |
+
['images/low00772.png'],
|
| 67 |
+
['images/low00723.png'],
|
| 68 |
+
['images/low00748.png'],
|
| 69 |
['images/1778_UHD_LL.JPG'],
|
| 70 |
['images/1791_UHD_LL.JPG']]
|
| 71 |
|
|
|
|
| 80 |
demo = gr.Interface(
|
| 81 |
fn=process_img,
|
| 82 |
inputs=[
|
| 83 |
+
gr.Image(type='pil', label='input', tool='editor'),
|
| 84 |
+
gr.Textbox(label="Image Filename", interactive=False)
|
| 85 |
],
|
| 86 |
outputs=[gr.Image(type='pil', label='output')],
|
| 87 |
title=title,
|
|
|
|
| 90 |
css=css
|
| 91 |
)
|
| 92 |
|
| 93 |
+
# Updating the filename in the input after selection
|
| 94 |
+
def update_filename(image):
|
| 95 |
+
# Retrieve the filename from the input image
|
| 96 |
+
if image:
|
| 97 |
+
filename = image.filename # Gradio automatically gives the file name
|
| 98 |
+
return filename
|
| 99 |
+
return ""
|
| 100 |
+
|
| 101 |
+
# Define the update logic for filename
|
| 102 |
+
demo.input_components[0].change(update_filename, inputs=demo.input_components[0], outputs=demo.input_components[1])
|
| 103 |
+
|
| 104 |
if __name__ == '__main__':
|
| 105 |
demo.launch()
|