Spaces:
Running
on
Zero
Running
on
Zero
Init model on GPU
Browse files- app.py +11 -3
- inference_gradio.py +21 -24
app.py
CHANGED
@@ -1,10 +1,10 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
import spaces
|
3 |
from inference_gradio import inference_one_image, model_init
|
4 |
|
5 |
MODEL_PATH = "./checkpoints/docres.pkl"
|
6 |
|
7 |
-
model = model_init(MODEL_PATH)
|
8 |
possible_tasks = [
|
9 |
"dewarping",
|
10 |
"deshadowing",
|
@@ -13,14 +13,22 @@ possible_tasks = [
|
|
13 |
"binarization",
|
14 |
]
|
15 |
|
16 |
-
@spaces.GPU
|
17 |
def run_tasks(image, tasks):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
18 |
bgr_image = image[..., ::-1].copy()
|
19 |
-
bgr_restored_image = inference_one_image(model, bgr_image, tasks)
|
20 |
if bgr_restored_image.ndim == 3:
|
21 |
rgb_image = bgr_restored_image[..., ::-1]
|
22 |
else:
|
23 |
rgb_image = bgr_restored_image
|
|
|
24 |
return rgb_image
|
25 |
|
26 |
|
|
|
1 |
+
import torch
|
2 |
import gradio as gr
|
3 |
import spaces
|
4 |
from inference_gradio import inference_one_image, model_init
|
5 |
|
6 |
MODEL_PATH = "./checkpoints/docres.pkl"
|
7 |
|
|
|
8 |
possible_tasks = [
|
9 |
"dewarping",
|
10 |
"deshadowing",
|
|
|
13 |
"binarization",
|
14 |
]
|
15 |
|
16 |
+
@spaces.GPU(duration=90)
|
17 |
def run_tasks(image, tasks):
|
18 |
+
|
19 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
20 |
+
|
21 |
+
# load model
|
22 |
+
model = model_init(MODEL_PATH, device)
|
23 |
+
|
24 |
+
# run inference
|
25 |
bgr_image = image[..., ::-1].copy()
|
26 |
+
bgr_restored_image = inference_one_image(model, bgr_image, tasks, device)
|
27 |
if bgr_restored_image.ndim == 3:
|
28 |
rgb_image = bgr_restored_image[..., ::-1]
|
29 |
else:
|
30 |
rgb_image = bgr_restored_image
|
31 |
+
|
32 |
return rgb_image
|
33 |
|
34 |
|
inference_gradio.py
CHANGED
@@ -14,9 +14,6 @@ sys.path.append("./data/MBD/")
|
|
14 |
from data.MBD.infer import net1_net2_infer_single_im
|
15 |
|
16 |
|
17 |
-
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
18 |
-
|
19 |
-
|
20 |
def dewarp_prompt(img):
|
21 |
mask = net1_net2_infer_single_im(img, "data/MBD/checkpoint/mbd.pkl")
|
22 |
base_coord = utils.getBasecoord(256, 256) / 256
|
@@ -122,7 +119,7 @@ def binarization_promptv2(img):
|
|
122 |
)
|
123 |
|
124 |
|
125 |
-
def dewarping(model, im_org):
|
126 |
INPUT_SIZE = 256
|
127 |
im_masked, prompt_org = dewarp_prompt(im_org.copy())
|
128 |
|
@@ -131,10 +128,10 @@ def dewarping(model, im_org):
|
|
131 |
im_masked = cv2.resize(im_masked, (INPUT_SIZE, INPUT_SIZE))
|
132 |
im_masked = im_masked / 255.0
|
133 |
im_masked = torch.from_numpy(im_masked.transpose(2, 0, 1)).unsqueeze(0)
|
134 |
-
im_masked = im_masked.float().to(
|
135 |
|
136 |
prompt = torch.from_numpy(prompt_org.transpose(2, 0, 1)).unsqueeze(0)
|
137 |
-
prompt = prompt.float().to(
|
138 |
|
139 |
in_im = torch.cat((im_masked, prompt), dim=1)
|
140 |
|
@@ -158,7 +155,7 @@ def dewarping(model, im_org):
|
|
158 |
return prompt_org[:, :, 0], prompt_org[:, :, 1], prompt_org[:, :, 2], out_im
|
159 |
|
160 |
|
161 |
-
def appearance(model, im_org):
|
162 |
MAX_SIZE = 1600
|
163 |
# obtain im and prompt
|
164 |
h, w = im_org.shape[:2]
|
@@ -176,7 +173,7 @@ def appearance(model, im_org):
|
|
176 |
in_im = torch.from_numpy(in_im.transpose(2, 0, 1)).unsqueeze(0)
|
177 |
|
178 |
# inference
|
179 |
-
in_im = in_im.half().to(
|
180 |
model = model.half()
|
181 |
with torch.no_grad():
|
182 |
pred = model(in_im)
|
@@ -198,7 +195,7 @@ def appearance(model, im_org):
|
|
198 |
return prompt[:, :, 0], prompt[:, :, 1], prompt[:, :, 2], out_im
|
199 |
|
200 |
|
201 |
-
def deshadowing(model, im_org):
|
202 |
MAX_SIZE = 1600
|
203 |
# obtain im and prompt
|
204 |
h, w = im_org.shape[:2]
|
@@ -216,7 +213,7 @@ def deshadowing(model, im_org):
|
|
216 |
in_im = torch.from_numpy(in_im.transpose(2, 0, 1)).unsqueeze(0)
|
217 |
|
218 |
# inference
|
219 |
-
in_im = in_im.half().to(
|
220 |
model = model.half()
|
221 |
with torch.no_grad():
|
222 |
pred = model(in_im)
|
@@ -238,16 +235,16 @@ def deshadowing(model, im_org):
|
|
238 |
return prompt[:, :, 0], prompt[:, :, 1], prompt[:, :, 2], out_im
|
239 |
|
240 |
|
241 |
-
def deblurring(model, im_org):
|
242 |
# setup image
|
243 |
in_im, padding_h, padding_w = stride_integral(im_org, 8)
|
244 |
prompt = deblur_prompt(in_im)
|
245 |
in_im = np.concatenate((in_im, prompt), -1)
|
246 |
in_im = in_im / 255.0
|
247 |
in_im = torch.from_numpy(in_im.transpose(2, 0, 1)).unsqueeze(0)
|
248 |
-
in_im = in_im.half().to(
|
249 |
# inference
|
250 |
-
model.to(
|
251 |
model.eval()
|
252 |
model = model.half()
|
253 |
with torch.no_grad():
|
@@ -260,7 +257,7 @@ def deblurring(model, im_org):
|
|
260 |
return prompt[:, :, 0], prompt[:, :, 1], prompt[:, :, 2], out_im
|
261 |
|
262 |
|
263 |
-
def binarization(model, im_org):
|
264 |
im, padding_h, padding_w = stride_integral(im_org, 8)
|
265 |
prompt = binarization_promptv2(im)
|
266 |
h, w = im.shape[:2]
|
@@ -268,7 +265,7 @@ def binarization(model, im_org):
|
|
268 |
|
269 |
in_im = in_im / 255.0
|
270 |
in_im = torch.from_numpy(in_im.transpose(2, 0, 1)).unsqueeze(0)
|
271 |
-
in_im = in_im.to(
|
272 |
model = model.half()
|
273 |
in_im = in_im.half()
|
274 |
with torch.no_grad():
|
@@ -283,7 +280,7 @@ def binarization(model, im_org):
|
|
283 |
return prompt[:, :, 0], prompt[:, :, 1], prompt[:, :, 2], out_im
|
284 |
|
285 |
|
286 |
-
def model_init(model_path):
|
287 |
# prepare model
|
288 |
model = restormer_arch.Restormer(
|
289 |
inp_channels=6,
|
@@ -298,7 +295,7 @@ def model_init(model_path):
|
|
298 |
dual_pixel_task=True,
|
299 |
)
|
300 |
|
301 |
-
if
|
302 |
state = convert_state_dict(
|
303 |
torch.load(model_path, map_location="cpu")["model_state"]
|
304 |
)
|
@@ -309,7 +306,7 @@ def model_init(model_path):
|
|
309 |
model.load_state_dict(state)
|
310 |
|
311 |
model.eval()
|
312 |
-
model = model.to(
|
313 |
return model
|
314 |
|
315 |
|
@@ -328,11 +325,11 @@ def resize(image, max_size):
|
|
328 |
return image
|
329 |
|
330 |
|
331 |
-
def inference_one_image(model, image, tasks):
|
332 |
# image should be in BGR format
|
333 |
|
334 |
if "dewarping" in tasks:
|
335 |
-
*_, image = dewarping(model, image)
|
336 |
|
337 |
# if only dewarping return here
|
338 |
if len(tasks) == 1 and "dewarping" in tasks:
|
@@ -341,12 +338,12 @@ def inference_one_image(model, image, tasks):
|
|
341 |
image = resize(image, 1536)
|
342 |
|
343 |
if "deshadowing" in tasks:
|
344 |
-
*_, image = deshadowing(model, image)
|
345 |
if "appearance" in tasks:
|
346 |
-
*_, image = appearance(model, image)
|
347 |
if "deblurring" in tasks:
|
348 |
-
*_, image = deblurring(model, image)
|
349 |
if "binarization" in tasks:
|
350 |
-
*_, image = binarization(model, image)
|
351 |
|
352 |
return image
|
|
|
14 |
from data.MBD.infer import net1_net2_infer_single_im
|
15 |
|
16 |
|
|
|
|
|
|
|
17 |
def dewarp_prompt(img):
|
18 |
mask = net1_net2_infer_single_im(img, "data/MBD/checkpoint/mbd.pkl")
|
19 |
base_coord = utils.getBasecoord(256, 256) / 256
|
|
|
119 |
)
|
120 |
|
121 |
|
122 |
+
def dewarping(model, im_org, device):
|
123 |
INPUT_SIZE = 256
|
124 |
im_masked, prompt_org = dewarp_prompt(im_org.copy())
|
125 |
|
|
|
128 |
im_masked = cv2.resize(im_masked, (INPUT_SIZE, INPUT_SIZE))
|
129 |
im_masked = im_masked / 255.0
|
130 |
im_masked = torch.from_numpy(im_masked.transpose(2, 0, 1)).unsqueeze(0)
|
131 |
+
im_masked = im_masked.float().to(device)
|
132 |
|
133 |
prompt = torch.from_numpy(prompt_org.transpose(2, 0, 1)).unsqueeze(0)
|
134 |
+
prompt = prompt.float().to(device)
|
135 |
|
136 |
in_im = torch.cat((im_masked, prompt), dim=1)
|
137 |
|
|
|
155 |
return prompt_org[:, :, 0], prompt_org[:, :, 1], prompt_org[:, :, 2], out_im
|
156 |
|
157 |
|
158 |
+
def appearance(model, im_org, device):
|
159 |
MAX_SIZE = 1600
|
160 |
# obtain im and prompt
|
161 |
h, w = im_org.shape[:2]
|
|
|
173 |
in_im = torch.from_numpy(in_im.transpose(2, 0, 1)).unsqueeze(0)
|
174 |
|
175 |
# inference
|
176 |
+
in_im = in_im.half().to(device)
|
177 |
model = model.half()
|
178 |
with torch.no_grad():
|
179 |
pred = model(in_im)
|
|
|
195 |
return prompt[:, :, 0], prompt[:, :, 1], prompt[:, :, 2], out_im
|
196 |
|
197 |
|
198 |
+
def deshadowing(model, im_org, device):
|
199 |
MAX_SIZE = 1600
|
200 |
# obtain im and prompt
|
201 |
h, w = im_org.shape[:2]
|
|
|
213 |
in_im = torch.from_numpy(in_im.transpose(2, 0, 1)).unsqueeze(0)
|
214 |
|
215 |
# inference
|
216 |
+
in_im = in_im.half().to(device)
|
217 |
model = model.half()
|
218 |
with torch.no_grad():
|
219 |
pred = model(in_im)
|
|
|
235 |
return prompt[:, :, 0], prompt[:, :, 1], prompt[:, :, 2], out_im
|
236 |
|
237 |
|
238 |
+
def deblurring(model, im_org, device):
|
239 |
# setup image
|
240 |
in_im, padding_h, padding_w = stride_integral(im_org, 8)
|
241 |
prompt = deblur_prompt(in_im)
|
242 |
in_im = np.concatenate((in_im, prompt), -1)
|
243 |
in_im = in_im / 255.0
|
244 |
in_im = torch.from_numpy(in_im.transpose(2, 0, 1)).unsqueeze(0)
|
245 |
+
in_im = in_im.half().to(device)
|
246 |
# inference
|
247 |
+
model.to(device)
|
248 |
model.eval()
|
249 |
model = model.half()
|
250 |
with torch.no_grad():
|
|
|
257 |
return prompt[:, :, 0], prompt[:, :, 1], prompt[:, :, 2], out_im
|
258 |
|
259 |
|
260 |
+
def binarization(model, im_org, device):
|
261 |
im, padding_h, padding_w = stride_integral(im_org, 8)
|
262 |
prompt = binarization_promptv2(im)
|
263 |
h, w = im.shape[:2]
|
|
|
265 |
|
266 |
in_im = in_im / 255.0
|
267 |
in_im = torch.from_numpy(in_im.transpose(2, 0, 1)).unsqueeze(0)
|
268 |
+
in_im = in_im.to(device)
|
269 |
model = model.half()
|
270 |
in_im = in_im.half()
|
271 |
with torch.no_grad():
|
|
|
280 |
return prompt[:, :, 0], prompt[:, :, 1], prompt[:, :, 2], out_im
|
281 |
|
282 |
|
283 |
+
def model_init(model_path, device):
|
284 |
# prepare model
|
285 |
model = restormer_arch.Restormer(
|
286 |
inp_channels=6,
|
|
|
295 |
dual_pixel_task=True,
|
296 |
)
|
297 |
|
298 |
+
if device == "cpu":
|
299 |
state = convert_state_dict(
|
300 |
torch.load(model_path, map_location="cpu")["model_state"]
|
301 |
)
|
|
|
306 |
model.load_state_dict(state)
|
307 |
|
308 |
model.eval()
|
309 |
+
model = model.to(device)
|
310 |
return model
|
311 |
|
312 |
|
|
|
325 |
return image
|
326 |
|
327 |
|
328 |
+
def inference_one_image(model, image, tasks, device):
|
329 |
# image should be in BGR format
|
330 |
|
331 |
if "dewarping" in tasks:
|
332 |
+
*_, image = dewarping(model, image, device)
|
333 |
|
334 |
# if only dewarping return here
|
335 |
if len(tasks) == 1 and "dewarping" in tasks:
|
|
|
338 |
image = resize(image, 1536)
|
339 |
|
340 |
if "deshadowing" in tasks:
|
341 |
+
*_, image = deshadowing(model, image, device)
|
342 |
if "appearance" in tasks:
|
343 |
+
*_, image = appearance(model, image, device)
|
344 |
if "deblurring" in tasks:
|
345 |
+
*_, image = deblurring(model, image, device)
|
346 |
if "binarization" in tasks:
|
347 |
+
*_, image = binarization(model, image, device)
|
348 |
|
349 |
return image
|