Spaces:
Sleeping
Sleeping
Upload 2 files
Browse files- app.py +56 -0
- requirements.txt +6 -0
app.py
ADDED
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gc
|
2 |
+
import torch
|
3 |
+
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
|
4 |
+
from craft_text_detector import Craft
|
5 |
+
from PIL import Image
|
6 |
+
import cv2
|
7 |
+
import time
|
8 |
+
import gradio as gr
|
9 |
+
|
10 |
+
# Force CPU usage, disable CUDA
|
11 |
+
torch.set_default_device('cpu')
|
12 |
+
craft = Craft(output_dir=None, crop_type="box", cuda=False)
|
13 |
+
|
14 |
+
# Load smaller model suitable for CPU
|
15 |
+
processor = TrOCRProcessor.from_pretrained('microsoft/trocr-small-handwritten')
|
16 |
+
model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-small-handwritten')
|
17 |
+
|
18 |
+
def recognize_handwritten(image):
|
19 |
+
start_time = time.time()
|
20 |
+
|
21 |
+
# Convert Gradio image to OpenCV format
|
22 |
+
image = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
|
23 |
+
result = craft.detect_text(image=image)
|
24 |
+
boxes = result["boxes"]
|
25 |
+
pil_image = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
|
26 |
+
texts = []
|
27 |
+
|
28 |
+
for box in boxes:
|
29 |
+
crop = pil_image.crop([box[0][0], box[0][1], box[2][0], box[2][1]])
|
30 |
+
pixel_values = processor(crop, return_tensors="pt").pixel_values
|
31 |
+
with torch.no_grad():
|
32 |
+
generated_ids = model.generate(pixel_values)
|
33 |
+
text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
34 |
+
texts.append(text)
|
35 |
+
|
36 |
+
text_data = " ".join(texts)
|
37 |
+
end_time = time.time()
|
38 |
+
time_difference = end_time - start_time
|
39 |
+
|
40 |
+
return f"Recognized text: {text_data}\nTime: {time_difference} seconds"
|
41 |
+
|
42 |
+
# Create Gradio interface
|
43 |
+
interface = gr.Interface(
|
44 |
+
fn=recognize_handwritten,
|
45 |
+
inputs=gr.Image(type="pil"),
|
46 |
+
outputs="text",
|
47 |
+
title="Handwritten Text Recognition",
|
48 |
+
description="Upload an image containing handwritten text to recognize it."
|
49 |
+
)
|
50 |
+
|
51 |
+
# Launch the app
|
52 |
+
interface.launch()
|
53 |
+
|
54 |
+
# Cleanup
|
55 |
+
craft.unload_craftnet_model()
|
56 |
+
gc.collect()
|
requirements.txt
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
torch
|
2 |
+
transformers
|
3 |
+
craft-text-detector
|
4 |
+
opencv-python
|
5 |
+
Pillow
|
6 |
+
gradio
|