Update app.py
Browse files
app.py
CHANGED
|
@@ -3,6 +3,8 @@ import torch
|
|
| 3 |
from PIL import Image
|
| 4 |
import numpy as np
|
| 5 |
from transformers import AutoProcessor, AutoModelForCausalLM
|
|
|
|
|
|
|
| 6 |
|
| 7 |
# Initialize Florence model
|
| 8 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
|
@@ -34,13 +36,12 @@ def generate_caption(image):
|
|
| 34 |
parsed_answer = florence_processor.post_process_generation(
|
| 35 |
generated_text,
|
| 36 |
task="<MORE_DETAILED_CAPTION>",
|
| 37 |
-
image_size=(image.width, image.height)
|
| 38 |
)
|
| 39 |
|
| 40 |
-
|
| 41 |
-
return prompt
|
| 42 |
|
| 43 |
-
# Streamlit
|
| 44 |
st.title("Florence 2 Caption Generator")
|
| 45 |
st.write("Upload an image to generate a caption:")
|
| 46 |
|
|
@@ -50,10 +51,29 @@ uploaded_image = st.file_uploader("Choose an Image", type=["jpg", "jpeg", "png"]
|
|
| 50 |
# If an image is uploaded
|
| 51 |
if uploaded_image is not None:
|
| 52 |
image = Image.open(uploaded_image)
|
| 53 |
-
st.image(image, caption="Uploaded Image", use_container_width=True)
|
| 54 |
-
|
| 55 |
# Generate caption when button is pressed
|
| 56 |
if st.button("Generate Caption"):
|
| 57 |
caption = generate_caption(image)
|
| 58 |
st.subheader("Generated Caption:")
|
| 59 |
-
st.write(caption)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
from PIL import Image
|
| 4 |
import numpy as np
|
| 5 |
from transformers import AutoProcessor, AutoModelForCausalLM
|
| 6 |
+
from io import BytesIO
|
| 7 |
+
import base64
|
| 8 |
|
| 9 |
# Initialize Florence model
|
| 10 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
|
|
|
| 36 |
parsed_answer = florence_processor.post_process_generation(
|
| 37 |
generated_text,
|
| 38 |
task="<MORE_DETAILED_CAPTION>",
|
| 39 |
+
image_size=(image.width, image.height)
|
| 40 |
)
|
| 41 |
|
| 42 |
+
return parsed_answer["<MORE_DETAILED_CAPTION>"]
|
|
|
|
| 43 |
|
| 44 |
+
# Streamlit UI
|
| 45 |
st.title("Florence 2 Caption Generator")
|
| 46 |
st.write("Upload an image to generate a caption:")
|
| 47 |
|
|
|
|
| 51 |
# If an image is uploaded
|
| 52 |
if uploaded_image is not None:
|
| 53 |
image = Image.open(uploaded_image)
|
| 54 |
+
st.image(image, caption="Uploaded Image", use_container_width=True)
|
| 55 |
+
|
| 56 |
# Generate caption when button is pressed
|
| 57 |
if st.button("Generate Caption"):
|
| 58 |
caption = generate_caption(image)
|
| 59 |
st.subheader("Generated Caption:")
|
| 60 |
+
st.write(caption)
|
| 61 |
+
|
| 62 |
+
# ✅ API Mode: Handle API Requests
|
| 63 |
+
st.experimental_set_query_params() # Ensure Streamlit can handle query params
|
| 64 |
+
|
| 65 |
+
def handle_api_request():
|
| 66 |
+
"""Handle API request by checking URL query parameters."""
|
| 67 |
+
query_params = st.experimental_get_query_params()
|
| 68 |
+
|
| 69 |
+
if "image" in query_params:
|
| 70 |
+
image_base64 = query_params["image"][0] # Get Base64-encoded image
|
| 71 |
+
image_bytes = BytesIO(base64.b64decode(image_base64))
|
| 72 |
+
image = Image.open(image_bytes)
|
| 73 |
+
|
| 74 |
+
caption = generate_caption(image)
|
| 75 |
+
st.json({"caption": caption}) # Return JSON response
|
| 76 |
+
|
| 77 |
+
# Check if API mode is enabled
|
| 78 |
+
if "image" in st.experimental_get_query_params():
|
| 79 |
+
handle_api_request()
|