Spaces:
Sleeping
Sleeping
quocanh944
commited on
Commit
β’
a1f1cc4
1
Parent(s):
93df5a7
fix url and app
Browse files
app.py
CHANGED
@@ -5,15 +5,6 @@ from PIL import Image
|
|
5 |
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-handwritten")
|
6 |
model = VisionEncoderDecoderModel.from_pretrained("quocanh944/tr-ocr")
|
7 |
|
8 |
-
# load image examples from the IAM database
|
9 |
-
urls = ['./images/a01-000u-00.png',
|
10 |
-
'./images/a01-000x-04.png',
|
11 |
-
'./images/a01-003-10.png']
|
12 |
-
|
13 |
-
for idx, url in enumerate(urls):
|
14 |
-
image = Image.open(url)
|
15 |
-
image.save(f"image_{idx}.png")
|
16 |
-
|
17 |
def process_image(image):
|
18 |
# prepare image
|
19 |
pixel_values = processor(image, return_tensors="pt").pixel_values
|
@@ -29,11 +20,11 @@ def process_image(image):
|
|
29 |
title = "Interactive demo: TrOCR"
|
30 |
description = "Demo for Microsoft's TrOCR, an encoder-decoder model consisting of an image Transformer encoder and a text Transformer decoder for state-of-the-art optical character recognition (OCR) on single-text line images. This particular model is fine-tuned on IAM, a dataset of annotated handwritten images. To use it, simply upload an image or use the example image below and click 'submit'. Results will show up in a few seconds."
|
31 |
article = "TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models | Github Repo"
|
32 |
-
examples =[["image_0.png"], ["image_1.png"], ["image_2.png"]]
|
33 |
|
34 |
iface = gr.Interface(fn=process_image,
|
35 |
-
inputs=gr.
|
36 |
-
outputs=gr.
|
37 |
title=title,
|
38 |
description=description,
|
39 |
article=article,
|
|
|
5 |
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-handwritten")
|
6 |
model = VisionEncoderDecoderModel.from_pretrained("quocanh944/tr-ocr")
|
7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
8 |
def process_image(image):
|
9 |
# prepare image
|
10 |
pixel_values = processor(image, return_tensors="pt").pixel_values
|
|
|
20 |
title = "Interactive demo: TrOCR"
|
21 |
description = "Demo for Microsoft's TrOCR, an encoder-decoder model consisting of an image Transformer encoder and a text Transformer decoder for state-of-the-art optical character recognition (OCR) on single-text line images. This particular model is fine-tuned on IAM, a dataset of annotated handwritten images. To use it, simply upload an image or use the example image below and click 'submit'. Results will show up in a few seconds."
|
22 |
article = "TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models | Github Repo"
|
23 |
+
examples =[["./images/image_0.png"], ["./images/image_1.png"], ["./images/image_2.png"], ["./images/image_3.png"]]
|
24 |
|
25 |
iface = gr.Interface(fn=process_image,
|
26 |
+
inputs=gr.Image(type="pil"),
|
27 |
+
outputs=gr.Textbox(),
|
28 |
title=title,
|
29 |
description=description,
|
30 |
article=article,
|
images/{a01-003-08.png β image_0.png}
RENAMED
File without changes
|
images/{a01-003u-01.png β image_1.png}
RENAMED
File without changes
|
images/{a01-003u-00.png β image_2.png}
RENAMED
File without changes
|
images/{a01-003-10.png β image_3.png}
RENAMED
File without changes
|