Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,34 +1,53 @@
|
|
1 |
import streamlit as st
|
2 |
-
import io
|
3 |
from PIL import Image
|
|
|
4 |
import torch
|
5 |
-
from transformers import AutoTokenizer, AutoModelForCausalLM
|
6 |
from easyocr import Reader
|
7 |
|
8 |
-
|
9 |
ocr_reader = Reader(['en'])
|
10 |
text_generator = AutoModelForCausalLM.from_pretrained("gpt2")
|
11 |
text_tokenizer = AutoTokenizer.from_pretrained("gpt2")
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
def extract_text(image):
|
13 |
return ocr_reader.readtext(image)
|
14 |
-
def explain_text(text):
|
15 |
-
input_ids = text_tokenizer.encode(text, return_tensors="pt")
|
16 |
-
explanation_ids = text_generator.generate(input_ids, max_length=100, num_return_sequences=1)
|
17 |
-
explanation = text_tokenizer.decode(explanation_ids[0], skip_special_tokens=True)
|
18 |
-
return explanation
|
19 |
|
20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
uploaded_file = st.file_uploader("Upload an image:")
|
22 |
|
|
|
23 |
if uploaded_file is not None:
|
24 |
image = Image.open(uploaded_file)
|
25 |
ocr_results = extract_text(image)
|
26 |
-
|
27 |
-
|
28 |
st.markdown("**Extracted text:**")
|
29 |
-
st.markdown(
|
30 |
|
31 |
-
st.markdown("**Explanation:**")
|
32 |
st.markdown(explanation)
|
33 |
|
34 |
else:
|
|
|
1 |
import streamlit as st
|
|
|
2 |
from PIL import Image
|
3 |
+
import io
|
4 |
import torch
|
5 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM, AutoProcessor, AutoModelForSeq2SeqLM
|
6 |
from easyocr import Reader
|
7 |
|
8 |
+
# Load the OCR model and text generation model
|
9 |
ocr_reader = Reader(['en'])
|
10 |
text_generator = AutoModelForCausalLM.from_pretrained("gpt2")
|
11 |
text_tokenizer = AutoTokenizer.from_pretrained("gpt2")
|
12 |
+
|
13 |
+
# Load the image captioning model
|
14 |
+
processor = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
|
15 |
+
caption_model = AutoModelForSeq2SeqLM.from_pretrained("Salesforce/blip-image-captioning-large")
|
16 |
+
|
17 |
+
# Define a function to extract text from an image using OCR
|
18 |
def extract_text(image):
|
19 |
return ocr_reader.readtext(image)
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
+
# Define a function to explain the extracted text using text generation
|
22 |
+
def explain_text(text, caption_model, processor):
|
23 |
+
# Extracted text
|
24 |
+
extracted_text = " ".join([res[1] for res in text])
|
25 |
+
|
26 |
+
# Generate an image caption using the image captioning model
|
27 |
+
inputs = processor(extracted_text, return_tensors="pt", padding="max_length", max_length=100, truncation=True)
|
28 |
+
input_ids = inputs["input_ids"]
|
29 |
+
caption = caption_model.generate(input_ids, max_length=50, num_return_sequences=1, no_repeat_ngram_size=2)
|
30 |
+
|
31 |
+
# Decode and return the generated caption
|
32 |
+
generated_caption = processor.decode(caption[0], skip_special_tokens=True)
|
33 |
+
return generated_caption
|
34 |
+
|
35 |
+
# Create a Streamlit layout
|
36 |
+
st.title("Text Extraction and Explanation")
|
37 |
+
|
38 |
+
# Allow users to upload an image
|
39 |
uploaded_file = st.file_uploader("Upload an image:")
|
40 |
|
41 |
+
# Extract text from the uploaded image and explain it
|
42 |
if uploaded_file is not None:
|
43 |
image = Image.open(uploaded_file)
|
44 |
ocr_results = extract_text(image)
|
45 |
+
explanation = explain_text(ocr_results, caption_model, processor)
|
46 |
+
|
47 |
st.markdown("**Extracted text:**")
|
48 |
+
st.markdown(" ".join([res[1] for res in ocr_results]))
|
49 |
|
50 |
+
st.markdown("**Explanation (Image Caption):**")
|
51 |
st.markdown(explanation)
|
52 |
|
53 |
else:
|