Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,11 +2,11 @@ import streamlit as st
|
|
| 2 |
from transformers import pipeline
|
| 3 |
from PIL import Image
|
| 4 |
|
| 5 |
-
# Page
|
| 6 |
-
st.set_page_config(page_title="TinyLLaVA
|
| 7 |
-
st.title("
|
| 8 |
|
| 9 |
-
# Load TinyLLaVA
|
| 10 |
pipe = pipeline(
|
| 11 |
task="image-to-text",
|
| 12 |
model="bczhou/tiny-llava-v1-hf",
|
|
@@ -14,21 +14,19 @@ pipe = pipeline(
|
|
| 14 |
device_map="cpu"
|
| 15 |
)
|
| 16 |
|
| 17 |
-
#
|
| 18 |
-
|
| 19 |
|
| 20 |
-
# Prompt
|
| 21 |
-
prompt = st.text_input("π¬
|
| 22 |
|
| 23 |
-
#
|
| 24 |
-
if
|
| 25 |
-
img = Image.open(
|
| 26 |
st.image(img, caption="Your Snapshot", use_column_width=True)
|
| 27 |
|
| 28 |
-
# Format query
|
| 29 |
query = f"USER: <image>\n{prompt}\nASSISTANT:"
|
| 30 |
-
|
| 31 |
-
with st.spinner("TinyLLaVA is thinking..."):
|
| 32 |
result = pipe(query, img)
|
| 33 |
answer = result[0]["generated_text"]
|
| 34 |
|
|
|
|
| 2 |
from transformers import pipeline
|
| 3 |
from PIL import Image
|
| 4 |
|
| 5 |
+
# Page config
|
| 6 |
+
st.set_page_config(page_title="TinyLLaVA Snapshot Q&A", layout="centered")
|
| 7 |
+
st.title("πΈ TinyLLaVA - Snapshot Q&A (Hugging Face Spaces)")
|
| 8 |
|
| 9 |
+
# Load TinyLLaVA pipeline
|
| 10 |
pipe = pipeline(
|
| 11 |
task="image-to-text",
|
| 12 |
model="bczhou/tiny-llava-v1-hf",
|
|
|
|
| 14 |
device_map="cpu"
|
| 15 |
)
|
| 16 |
|
| 17 |
+
# Always visible camera input
|
| 18 |
+
image = st.camera_input("π· Take a snapshot")
|
| 19 |
|
| 20 |
+
# Prompt text input
|
| 21 |
+
prompt = st.text_input("π¬ Your question:", value="Describe this scene.")
|
| 22 |
|
| 23 |
+
# If both present, run inference
|
| 24 |
+
if image is not None and prompt:
|
| 25 |
+
img = Image.open(image).convert("RGB")
|
| 26 |
st.image(img, caption="Your Snapshot", use_column_width=True)
|
| 27 |
|
|
|
|
| 28 |
query = f"USER: <image>\n{prompt}\nASSISTANT:"
|
| 29 |
+
with st.spinner("TinyLLaVA is generating..."):
|
|
|
|
| 30 |
result = pipe(query, img)
|
| 31 |
answer = result[0]["generated_text"]
|
| 32 |
|