Update app.py
Browse files
app.py
CHANGED
@@ -1,4 +1,3 @@
|
|
1 |
-
from huggingface_hub import from_pretrained_fastai
|
2 |
import gradio as gr
|
3 |
from fastai.basics import *
|
4 |
from fastai.vision import models
|
@@ -6,65 +5,12 @@ from fastai.vision.all import *
|
|
6 |
from fastai.metrics import *
|
7 |
from fastai.data.all import *
|
8 |
from fastai.callback import *
|
9 |
-
|
10 |
-
|
11 |
-
from pathlib import Path
|
12 |
-
import random
|
13 |
import PIL
|
14 |
-
import torchvision.transforms as
|
15 |
-
import torch
|
16 |
-
from albumentations import *
|
17 |
-
|
18 |
-
def get_y_fn (x):
|
19 |
-
return Path(str(x).replace("Images","Labels").replace("color","gt").replace(".jpg",".png"))
|
20 |
-
|
21 |
-
def ParentSplitter(x):
|
22 |
-
return Path(x).parent.name==test_name
|
23 |
-
|
24 |
-
class SegmentationAlbumentationsTransform(ItemTransform):
|
25 |
-
split_idx = 0
|
26 |
-
|
27 |
-
def __init__(self, aug):
|
28 |
-
self.aug = aug
|
29 |
-
|
30 |
-
def encodes(self, x):
|
31 |
-
img,mask = x
|
32 |
-
aug = self.aug(image=np.array(img), mask=np.array(mask))
|
33 |
-
return PILImage.create(aug["image"]), PILMask.create(aug["mask"])
|
34 |
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
transformPipeline=SegmentationAlbumentationsTransform(transforms)
|
39 |
-
|
40 |
-
class TargetMaskConvertTransform(ItemTransform):
|
41 |
-
def __init__(self):
|
42 |
-
pass
|
43 |
-
def encodes(self, x):
|
44 |
-
img,mask = x
|
45 |
-
|
46 |
-
#Convert to array
|
47 |
-
mask = np.array(mask)
|
48 |
-
|
49 |
-
# Aquí definimos cada clase en la máscara
|
50 |
-
# uva:
|
51 |
-
mask[mask==255]=1
|
52 |
-
# hojas:
|
53 |
-
mask[mask==150]=2
|
54 |
-
# conductores:
|
55 |
-
mask[mask==76]=3
|
56 |
-
mask[mask==74]=3
|
57 |
-
# madera:
|
58 |
-
mask[mask==29]=4
|
59 |
-
mask[mask==25]=4
|
60 |
-
|
61 |
-
# Back to PILMask
|
62 |
-
mask = PILMask.create(mask)
|
63 |
-
return img, mask
|
64 |
-
|
65 |
-
repo_id = "paascorb/practica3_Segmentation"
|
66 |
-
|
67 |
-
learner = from_pretrained_fastai(repo_id)
|
68 |
|
69 |
def transform_image(image):
|
70 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
@@ -79,8 +25,7 @@ def predict(img):
|
|
79 |
img = PIL.Image.fromarray(img, "RGB")
|
80 |
image = transformss.Resize((480,640))(img)
|
81 |
tensor = transform_image(image=image)
|
82 |
-
|
83 |
-
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
84 |
learner.to(device)
|
85 |
with torch.no_grad():
|
86 |
outputs = learner(tensor)
|
|
|
|
|
1 |
import gradio as gr
|
2 |
from fastai.basics import *
|
3 |
from fastai.vision import models
|
|
|
5 |
from fastai.metrics import *
|
6 |
from fastai.data.all import *
|
7 |
from fastai.callback import *
|
|
|
|
|
|
|
|
|
8 |
import PIL
|
9 |
+
import torchvision.transforms as transforms
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
12 |
+
model = torch.jit.load("unet.pth")
|
13 |
+
model = model.cpu()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
def transform_image(image):
|
16 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
|
25 |
img = PIL.Image.fromarray(img, "RGB")
|
26 |
image = transformss.Resize((480,640))(img)
|
27 |
tensor = transform_image(image=image)
|
28 |
+
|
|
|
29 |
learner.to(device)
|
30 |
with torch.no_grad():
|
31 |
outputs = learner(tensor)
|