Spaces:
Sleeping
Sleeping
wjbmattingly
commited on
Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,6 @@
|
|
1 |
import gradio as gr
|
2 |
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
|
3 |
-
import
|
4 |
-
from PIL import Image
|
5 |
|
6 |
# Dictionary of model names and their corresponding HuggingFace model IDs
|
7 |
MODEL_OPTIONS = {
|
@@ -18,49 +17,57 @@ MODEL_OPTIONS = {
|
|
18 |
"Medieval Print": "medieval-data/trocr-medieval-print"
|
19 |
}
|
20 |
|
21 |
-
#
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
for idx, url in enumerate(urls):
|
27 |
-
image = Image.open(requests.get(url, stream=True).raw)
|
28 |
-
image.save(f"image_{idx}.png")
|
29 |
|
30 |
def load_model(model_name):
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
|
36 |
def process_image(image, model_name):
|
37 |
processor, model = load_model(model_name)
|
38 |
|
39 |
-
#
|
40 |
pixel_values = processor(image, return_tensors="pt").pixel_values
|
41 |
|
42 |
-
#
|
43 |
-
|
|
|
44 |
|
45 |
-
#
|
|
|
|
|
|
|
|
|
46 |
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
47 |
return generated_text
|
48 |
|
49 |
-
|
50 |
-
description = "Demo for the Medieval TrOCR HTR Models."
|
51 |
-
|
52 |
iface = gr.Interface(
|
53 |
fn=process_image,
|
54 |
inputs=[
|
55 |
-
gr.Image(type="pil"),
|
56 |
-
gr.Dropdown(choices=list(MODEL_OPTIONS.keys()), label="Select Model")
|
57 |
],
|
58 |
-
outputs=gr.Textbox(),
|
59 |
-
title=
|
60 |
-
description=
|
61 |
examples=[
|
62 |
-
["
|
63 |
]
|
64 |
)
|
65 |
|
66 |
-
iface.launch(
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
|
3 |
+
import torch
|
|
|
4 |
|
5 |
# Dictionary of model names and their corresponding HuggingFace model IDs
|
6 |
MODEL_OPTIONS = {
|
|
|
17 |
"Medieval Print": "medieval-data/trocr-medieval-print"
|
18 |
}
|
19 |
|
20 |
+
# Global variables to store the current model and processor
|
21 |
+
current_model = None
|
22 |
+
current_processor = None
|
23 |
+
current_model_name = None
|
|
|
|
|
|
|
|
|
24 |
|
25 |
def load_model(model_name):
|
26 |
+
global current_model, current_processor, current_model_name
|
27 |
+
|
28 |
+
if model_name != current_model_name:
|
29 |
+
model_id = MODEL_OPTIONS[model_name]
|
30 |
+
current_processor = TrOCRProcessor.from_pretrained(model_id)
|
31 |
+
current_model = VisionEncoderDecoderModel.from_pretrained(model_id)
|
32 |
+
current_model_name = model_name
|
33 |
+
|
34 |
+
# Move model to GPU if available
|
35 |
+
if torch.cuda.is_available():
|
36 |
+
current_model = current_model.to('cuda')
|
37 |
+
|
38 |
+
return current_processor, current_model
|
39 |
|
40 |
def process_image(image, model_name):
|
41 |
processor, model = load_model(model_name)
|
42 |
|
43 |
+
# Prepare image
|
44 |
pixel_values = processor(image, return_tensors="pt").pixel_values
|
45 |
|
46 |
+
# Move input to GPU if model is on GPU
|
47 |
+
if next(model.parameters()).is_cuda:
|
48 |
+
pixel_values = pixel_values.to('cuda')
|
49 |
|
50 |
+
# Generate (no beam search)
|
51 |
+
with torch.no_grad():
|
52 |
+
generated_ids = model.generate(pixel_values)
|
53 |
+
|
54 |
+
# Decode
|
55 |
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
56 |
return generated_text
|
57 |
|
58 |
+
# Gradio interface
|
|
|
|
|
59 |
iface = gr.Interface(
|
60 |
fn=process_image,
|
61 |
inputs=[
|
62 |
+
gr.Image(type="pil", label="Input Image"),
|
63 |
+
gr.Dropdown(choices=list(MODEL_OPTIONS.keys()), label="Select Model", value="Medieval Base")
|
64 |
],
|
65 |
+
outputs=gr.Textbox(label="Transcription"),
|
66 |
+
title="Medieval TrOCR Model Switcher",
|
67 |
+
description="Upload an image of medieval text and select a model to transcribe it.",
|
68 |
examples=[
|
69 |
+
["https://huggingface.co/medieval-data/trocr-medieval-base/resolve/main/images/caroline-1.png", "Medieval Latin Caroline"]
|
70 |
]
|
71 |
)
|
72 |
|
73 |
+
iface.launch()
|