Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,48 +1,33 @@
|
|
1 |
import gradio as gr
|
2 |
-
import torch
|
3 |
-
import numpy as np
|
4 |
-
from PIL import Image
|
5 |
-
from transformers import AutoProcessor, AutoModelForVision2Seq
|
6 |
-
|
7 |
-
# Load the model & processor
|
8 |
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
|
|
|
9 |
|
|
|
10 |
model_name = "microsoft/trocr-large-handwritten"
|
11 |
processor = TrOCRProcessor.from_pretrained(model_name)
|
12 |
model = VisionEncoderDecoderModel.from_pretrained(model_name)
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
# Function to extract handwritten text
|
18 |
-
def extract_text(image):
|
19 |
-
try:
|
20 |
-
# Convert input to PIL Image
|
21 |
-
if isinstance(image, np.ndarray):
|
22 |
-
if len(image.shape) == 2: # If grayscale (H, W), add channels
|
23 |
-
image = np.stack([image] * 3, axis=-1)
|
24 |
-
image = Image.fromarray(image)
|
25 |
-
else:
|
26 |
-
image = Image.open(image).convert("RGB")
|
27 |
-
|
28 |
-
# Process image through model
|
29 |
-
pixel_values = processor(images=image, return_tensors="pt").pixel_values
|
30 |
-
generated_ids = model.generate(pixel_values)
|
31 |
-
extracted_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
32 |
-
|
33 |
-
return extracted_text
|
34 |
|
35 |
-
|
36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
|
38 |
-
# Gradio
|
39 |
iface = gr.Interface(
|
40 |
-
fn=
|
41 |
-
inputs="
|
42 |
outputs="text",
|
43 |
-
title="Handwritten
|
44 |
-
description="Upload a handwritten
|
45 |
)
|
46 |
|
47 |
-
#
|
48 |
iface.launch()
|
|
|
1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
|
3 |
+
from PIL import Image
|
4 |
|
5 |
+
# Load the model and processor
|
6 |
model_name = "microsoft/trocr-large-handwritten"
|
7 |
processor = TrOCRProcessor.from_pretrained(model_name)
|
8 |
model = VisionEncoderDecoderModel.from_pretrained(model_name)
|
9 |
|
10 |
+
def ocr_recognition(image):
|
11 |
+
# Open the image
|
12 |
+
image = Image.open(image).convert("RGB")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
+
# Process the image and generate text
|
15 |
+
pixel_values = processor(image, return_tensors="pt").pixel_values
|
16 |
+
generated_ids = model.generate(pixel_values)
|
17 |
+
|
18 |
+
# Decode the output text
|
19 |
+
text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
20 |
+
|
21 |
+
return text
|
22 |
|
23 |
+
# Create Gradio interface
|
24 |
iface = gr.Interface(
|
25 |
+
fn=ocr_recognition,
|
26 |
+
inputs=gr.Image(type="pil"), # Ensures PIL image input
|
27 |
outputs="text",
|
28 |
+
title="Handwritten OCR Extraction",
|
29 |
+
description="Upload a handwritten image to extract text using TrOCR."
|
30 |
)
|
31 |
|
32 |
+
# Launch the Gradio app
|
33 |
iface.launch()
|