Spaces:
Sleeping
Sleeping
add mask to rgb label2color image
Browse files- app.py +21 -7
- requirements.txt +1 -0
app.py
CHANGED
@@ -1,4 +1,6 @@
|
|
1 |
import gradio as gr
|
|
|
|
|
2 |
import torch
|
3 |
from torch import nn
|
4 |
from transformers import (SegformerFeatureExtractor,
|
@@ -23,23 +25,35 @@ def upscale_logits(logit_outputs, size):
|
|
23 |
align_corners=False
|
24 |
)
|
25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
def query_image(img):
|
27 |
"""Función para generar predicciones a la escala origina"""
|
28 |
inputs = preprocessor(images=img, return_tensors="pt")
|
29 |
with torch.no_grad():
|
30 |
-
#preds = model(inputs.unsqueeze(0).to(device))["logits"]
|
31 |
preds = model(**inputs)["logits"]
|
32 |
-
preds_upscale = upscale_logits(preds,
|
33 |
predict_label = torch.argmax(preds_upscale, dim=1).to(device)
|
34 |
-
|
35 |
-
|
36 |
|
37 |
-
def visualize_instance_seg_mask(mask):
|
38 |
-
return mask
|
39 |
|
40 |
demo = gr.Interface(
|
41 |
query_image,
|
42 |
-
inputs=[gr.Image()],
|
43 |
outputs="image",
|
44 |
title="SegFormer Model for rock glacier image segmentation"
|
45 |
)
|
|
|
1 |
import gradio as gr
|
2 |
+
import random
|
3 |
+
import numpy as np
|
4 |
import torch
|
5 |
from torch import nn
|
6 |
from transformers import (SegformerFeatureExtractor,
|
|
|
25 |
align_corners=False
|
26 |
)
|
27 |
|
28 |
+
|
29 |
+
def visualize_instance_seg_mask(mask):
|
30 |
+
"""Agrega colores RGB a cada una de las clases en la mask"""
|
31 |
+
image = np.zeros((mask.shape[0], mask.shape[1], 3))
|
32 |
+
labels = np.unique(mask)
|
33 |
+
label2color = {label: (random.randint(0, 1),
|
34 |
+
random.randint(0, 255),
|
35 |
+
random.randint(0, 255)) for label in labels}
|
36 |
+
for i in range(image.shape[0]):
|
37 |
+
for j in range(image.shape[1]):
|
38 |
+
image[i, j, :] = label2color[mask[i, j]]
|
39 |
+
image = image / 255
|
40 |
+
return image
|
41 |
+
|
42 |
+
|
43 |
def query_image(img):
|
44 |
"""Función para generar predicciones a la escala origina"""
|
45 |
inputs = preprocessor(images=img, return_tensors="pt")
|
46 |
with torch.no_grad():
|
|
|
47 |
preds = model(**inputs)["logits"]
|
48 |
+
preds_upscale = upscale_logits(preds, preds.shape[2])
|
49 |
predict_label = torch.argmax(preds_upscale, dim=1).to(device)
|
50 |
+
result = predict_label[0,:,:].detach().cpu().numpy()
|
51 |
+
return visualize_instance_seg_mask(result)
|
52 |
|
|
|
|
|
53 |
|
54 |
demo = gr.Interface(
|
55 |
query_image,
|
56 |
+
inputs=[gr.Image(type="pil")],
|
57 |
outputs="image",
|
58 |
title="SegFormer Model for rock glacier image segmentation"
|
59 |
)
|
requirements.txt
CHANGED
@@ -1,2 +1,3 @@
|
|
1 |
torch
|
2 |
transformers
|
|
|
|
1 |
torch
|
2 |
transformers
|
3 |
+
numpy
|