Spaces:
Runtime error
Runtime error
finish
Browse files
app.py
CHANGED
@@ -6,14 +6,16 @@ import numpy as np
|
|
6 |
from PIL import Image
|
7 |
import tensorflow as tf
|
8 |
from transformers import SegformerFeatureExtractor, TFSegformerForSemanticSegmentation
|
|
|
9 |
|
10 |
feature_extractor = SegformerFeatureExtractor.from_pretrained(
|
11 |
"nvidia/segformer-b2-finetuned-cityscapes-1024-1024"
|
12 |
)
|
13 |
-
|
14 |
"nvidia/segformer-b2-finetuned-cityscapes-1024-1024"
|
15 |
)
|
16 |
|
|
|
17 |
|
18 |
def ade_palette():
|
19 |
"""ADE20K palette that maps each class to RGB values."""
|
@@ -82,7 +84,7 @@ def sepia(input_img):
|
|
82 |
input_img = Image.fromarray(input_img)
|
83 |
|
84 |
inputs = feature_extractor(images=input_img, return_tensors="tf")
|
85 |
-
outputs =
|
86 |
logits = outputs.logits
|
87 |
|
88 |
logits = tf.transpose(logits, [0, 2, 3, 1])
|
@@ -105,11 +107,12 @@ def sepia(input_img):
|
|
105 |
return fig
|
106 |
|
107 |
|
108 |
-
def
|
109 |
input_img = Image.fromarray(input_img)
|
110 |
|
|
|
111 |
inputs = feature_extractor(images=input_img, return_tensors="tf")
|
112 |
-
outputs =
|
113 |
logits = outputs.logits
|
114 |
|
115 |
logits = tf.transpose(logits, [0, 2, 3, 1])
|
@@ -118,21 +121,25 @@ def with_labels(input_img):
|
|
118 |
)
|
119 |
seg = tf.math.argmax(logits, axis=-1)[0]
|
120 |
|
121 |
-
|
122 |
-
|
123 |
-
)
|
124 |
-
|
125 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
126 |
|
127 |
-
pred_img = np.array(input_img) * 0.5 + color_seg * 0.5
|
128 |
-
pred_img = pred_img.astype(np.uint8)
|
129 |
|
130 |
-
return input_img, pred_img, labels_list
|
131 |
|
132 |
|
133 |
-
demo = gr.Interface(fn=
|
134 |
inputs=gr.Image(shape=(1024, 1024)),
|
135 |
-
outputs=["image",
|
136 |
examples=["city-1.jpg", "city-2.jpg", "city-3.jpg", "city-4.jpg", "city-5.jpg"],
|
137 |
allow_flagging='never')
|
138 |
|
|
|
6 |
from PIL import Image
|
7 |
import tensorflow as tf
|
8 |
from transformers import SegformerFeatureExtractor, TFSegformerForSemanticSegmentation
|
9 |
+
from transformers import BigBirdForImageCaptioning
|
10 |
|
11 |
feature_extractor = SegformerFeatureExtractor.from_pretrained(
|
12 |
"nvidia/segformer-b2-finetuned-cityscapes-1024-1024"
|
13 |
)
|
14 |
+
seg_model = TFSegformerForSemanticSegmentation.from_pretrained(
|
15 |
"nvidia/segformer-b2-finetuned-cityscapes-1024-1024"
|
16 |
)
|
17 |
|
18 |
+
caption_model = BigBirdForImageCaptioning.from_pretrained("bigbird/image-captioning-base")
|
19 |
|
20 |
def ade_palette():
|
21 |
"""ADE20K palette that maps each class to RGB values."""
|
|
|
84 |
input_img = Image.fromarray(input_img)
|
85 |
|
86 |
inputs = feature_extractor(images=input_img, return_tensors="tf")
|
87 |
+
outputs = seg_model(**inputs)
|
88 |
logits = outputs.logits
|
89 |
|
90 |
logits = tf.transpose(logits, [0, 2, 3, 1])
|
|
|
107 |
return fig
|
108 |
|
109 |
|
110 |
+
def segment_and_caption(input_img):
|
111 |
input_img = Image.fromarray(input_img)
|
112 |
|
113 |
+
# ์ธ๊ทธ๋ฉํ
์ด์
์ํ
|
114 |
inputs = feature_extractor(images=input_img, return_tensors="tf")
|
115 |
+
outputs = seg_model(**inputs)
|
116 |
logits = outputs.logits
|
117 |
|
118 |
logits = tf.transpose(logits, [0, 2, 3, 1])
|
|
|
121 |
)
|
122 |
seg = tf.math.argmax(logits, axis=-1)[0]
|
123 |
|
124 |
+
# ์ธ๊ทธ๋ฉํ
์ด์
๊ฒฐ๊ณผ๋ฅผ ํ
์คํธ๋ก ๋ณํ
|
125 |
+
seg_text = ""
|
126 |
+
for label, label_name in enumerate(labels_list):
|
127 |
+
count = np.sum(seg.numpy() == label)
|
128 |
+
seg_text += f"{label_name}: {count} pixels\n"
|
129 |
+
|
130 |
+
# ์ด๋ฏธ์ง ์บก์
์์ฑ
|
131 |
+
caption = caption_model.generate(input_img, max_length=20, num_return_sequences=1, return_dict_in_generate=True)
|
132 |
+
caption_text = caption[0]['text']
|
133 |
+
|
134 |
+
# ์ธ๊ทธ๋ฉํ
์ด์
๊ฒฐ๊ณผ์ ์บก์
์ ๋ฐํ
|
135 |
+
return input_img, seg_text, caption_text
|
136 |
|
|
|
|
|
137 |
|
|
|
138 |
|
139 |
|
140 |
+
demo = gr.Interface(fn=segment_and_caption,
|
141 |
inputs=gr.Image(shape=(1024, 1024)),
|
142 |
+
outputs=["image","text", "text"],
|
143 |
examples=["city-1.jpg", "city-2.jpg", "city-3.jpg", "city-4.jpg", "city-5.jpg"],
|
144 |
allow_flagging='never')
|
145 |
|