Upload 3 files
Browse files- app.py +99 -0
- image.PNG +0 -0
- requirements.txt +3 -0
app.py
ADDED
|
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import streamlit as st
|
| 3 |
+
import google.generativeai as genai
|
| 4 |
+
from PIL import Image
|
| 5 |
+
import base64
|
| 6 |
+
|
| 7 |
+
# Function to set the Google API key
|
| 8 |
+
def set_api_key(api_key):
|
| 9 |
+
os.environ["GOOGLE_API_KEY"] = api_key
|
| 10 |
+
genai.configure(api_key=api_key)
|
| 11 |
+
|
| 12 |
+
# Function to get description from the image
|
| 13 |
+
def get_image_description(image, user_prompt=None):
|
| 14 |
+
model = genai.GenerativeModel(model_name="gemini-1.5-flash")
|
| 15 |
+
prompt = user_prompt if user_prompt else "What is in this photo?"
|
| 16 |
+
response = model.generate_content([prompt], images=[image])
|
| 17 |
+
return response.text
|
| 18 |
+
|
| 19 |
+
# Helper function to convert image to base64
|
| 20 |
+
def get_base64_of_image(image_path):
|
| 21 |
+
with open(image_path, "rb") as image_file:
|
| 22 |
+
base64_str = base64.b64encode(image_file.read()).decode()
|
| 23 |
+
return base64_str
|
| 24 |
+
|
| 25 |
+
def main():
|
| 26 |
+
st.set_page_config(page_title="Talk to Image", layout="wide")
|
| 27 |
+
|
| 28 |
+
st.markdown(
|
| 29 |
+
f"""
|
| 30 |
+
<style>
|
| 31 |
+
.stApp {{
|
| 32 |
+
background: url(data:image/png;base64,{get_base64_of_image('image.png')});
|
| 33 |
+
background-size: cover
|
| 34 |
+
}}
|
| 35 |
+
</style>
|
| 36 |
+
""",
|
| 37 |
+
unsafe_allow_html=True
|
| 38 |
+
)
|
| 39 |
+
|
| 40 |
+
st.title("Image Description using Gemini!")
|
| 41 |
+
|
| 42 |
+
st.subheader("Upload your Image File")
|
| 43 |
+
uploaded_file = st.file_uploader("Upload your image file", type=["png", "jpg", "jpeg"])
|
| 44 |
+
|
| 45 |
+
st.sidebar.header("Configuration")
|
| 46 |
+
api_key = st.sidebar.text_input("Google API Key:", type="password")
|
| 47 |
+
|
| 48 |
+
if st.sidebar.button("Set API Key"):
|
| 49 |
+
if api_key:
|
| 50 |
+
set_api_key(api_key)
|
| 51 |
+
st.sidebar.success("API Key set successfully.")
|
| 52 |
+
else:
|
| 53 |
+
st.sidebar.error("Please enter a valid API Key.")
|
| 54 |
+
|
| 55 |
+
prompt = st.text_input("Enter your prompt (optional):", value="What is in this photo?")
|
| 56 |
+
|
| 57 |
+
if st.button("Describe Image") and uploaded_file:
|
| 58 |
+
if "GOOGLE_API_KEY" in os.environ and os.environ["GOOGLE_API_KEY"]:
|
| 59 |
+
try:
|
| 60 |
+
image = Image.open(uploaded_file)
|
| 61 |
+
st.image(image, caption="Uploaded Image", use_column_width=True)
|
| 62 |
+
|
| 63 |
+
with st.spinner("Generating description..."):
|
| 64 |
+
description = get_image_description(image, prompt)
|
| 65 |
+
st.write("**Description:** ", description)
|
| 66 |
+
except Exception as e:
|
| 67 |
+
st.error(f"An error occurred: {e}")
|
| 68 |
+
else:
|
| 69 |
+
st.warning("Please enter your Google API key and click 'Set API Key'.")
|
| 70 |
+
|
| 71 |
+
st.markdown("---")
|
| 72 |
+
st.write("Happy to Connect:")
|
| 73 |
+
kaggle, linkedin, google_scholar, youtube, github = st.columns(5)
|
| 74 |
+
|
| 75 |
+
image_urls = {
|
| 76 |
+
"kaggle": "https://www.kaggle.com/static/images/site-logo.svg",
|
| 77 |
+
"linkedin": "https://upload.wikimedia.org/wikipedia/commons/thumb/c/ca/LinkedIn_logo_initials.png/600px-LinkedIn_logo_initials.png",
|
| 78 |
+
"google_scholar": "https://upload.wikimedia.org/wikipedia/commons/thumb/c/c7/Google_Scholar_logo.svg/768px-Google_Scholar_logo.svg.png",
|
| 79 |
+
"youtube": "https://upload.wikimedia.org/wikipedia/commons/thumb/7/72/YouTube_social_white_square_%282017%29.svg/640px-YouTube_social_white_square_%282017%29.svg.png",
|
| 80 |
+
"github": "https://github.githubassets.com/assets/GitHub-Mark-ea2971cee799.png"
|
| 81 |
+
}
|
| 82 |
+
|
| 83 |
+
social_links = {
|
| 84 |
+
"kaggle": "https://www.kaggle.com/muhammadimran112233",
|
| 85 |
+
"linkedin": "https://www.linkedin.com/in/muhammad-imran-zaman",
|
| 86 |
+
"google_scholar": "https://scholar.google.com/citations?user=ulVFpy8AAAAJ&hl=en",
|
| 87 |
+
"youtube": "https://www.youtube.com/@consolioo",
|
| 88 |
+
"github": "https://github.com/Imran-ml"
|
| 89 |
+
}
|
| 90 |
+
|
| 91 |
+
kaggle.markdown(f'<a href="{social_links["kaggle"]}"><img src="{image_urls["kaggle"]}" width="50" height="50"></a>', unsafe_allow_html=True)
|
| 92 |
+
linkedin.markdown(f'<a href="{social_links["linkedin"]}"><img src="{image_urls["linkedin"]}" width="50" height="50"></a>', unsafe_allow_html=True)
|
| 93 |
+
google_scholar.markdown(f'<a href="{social_links["google_scholar"]}"><img src="{image_urls["google_scholar"]}" width="50" height="50"></a>', unsafe_allow_html=True)
|
| 94 |
+
youtube.markdown(f'<a href="{social_links["youtube"]}"><img src="{image_urls["youtube"]}" width="50" height="50"></a>', unsafe_allow_html=True)
|
| 95 |
+
github.markdown(f'<a href="{social_links["github"]}"><img src="{image_urls["github"]}" width="50" height="50"></a>', unsafe_allow_html=True)
|
| 96 |
+
st.markdown("---")
|
| 97 |
+
|
| 98 |
+
if __name__ == "__main__":
|
| 99 |
+
main()
|
image.PNG
ADDED
|
|
requirements.txt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
streamlit
|
| 2 |
+
google-generativeai
|
| 3 |
+
Pillow
|