Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,127 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import cv2
|
3 |
+
import requests
|
4 |
+
import base64
|
5 |
+
import tempfile
|
6 |
+
import os
|
7 |
+
import time
|
8 |
+
from typing import Generator, Tuple
|
9 |
+
|
10 |
+
# --------------------------
|
11 |
+
# Configuration
|
12 |
+
# --------------------------
|
13 |
+
API_KEY = os.getenv("GEMINI_API_KEY") # Fetch API key from Hugging Face secrets
|
14 |
+
API_URL = f"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?key={API_KEY}"
|
15 |
+
SYSTEM_PROMPT = '''
|
16 |
+
You are a next-generation AI-driven military surveillance officer...
|
17 |
+
(Same as before)
|
18 |
+
'''
|
19 |
+
|
20 |
+
# --------------------------
|
21 |
+
# Gemini API Client
|
22 |
+
# --------------------------
|
23 |
+
class GeminiClient:
|
24 |
+
def __init__(self, api_key: str, api_url: str):
|
25 |
+
self.api_key = api_key
|
26 |
+
self.api_url = api_url
|
27 |
+
self.session = requests.Session()
|
28 |
+
self.timeout = 30
|
29 |
+
|
30 |
+
def analyze_frame(self, frame_b64: str, timestamp: int) -> str:
|
31 |
+
"""Send frame to Gemini API for analysis."""
|
32 |
+
payload = {
|
33 |
+
"contents": [{
|
34 |
+
"parts": [
|
35 |
+
{"text": f"Analyze this battlefield image from {timestamp} seconds:"},
|
36 |
+
{"inline_data": {"mime_type": "image/jpeg", "data": frame_b64}}
|
37 |
+
]
|
38 |
+
}],
|
39 |
+
"systemInstruction": {
|
40 |
+
"parts": [{"text": SYSTEM_PROMPT}]
|
41 |
+
}
|
42 |
+
}
|
43 |
+
|
44 |
+
try:
|
45 |
+
response = self.session.post(
|
46 |
+
self.api_url,
|
47 |
+
json=payload,
|
48 |
+
timeout=self.timeout,
|
49 |
+
headers={"Content-Type": "application/json"}
|
50 |
+
)
|
51 |
+
response.raise_for_status()
|
52 |
+
return self._parse_response(response.json())
|
53 |
+
except requests.exceptions.RequestException as e:
|
54 |
+
return f"Analysis error: {str(e)}"
|
55 |
+
|
56 |
+
@staticmethod
|
57 |
+
def _parse_response(response: dict) -> str:
|
58 |
+
"""Extract response text from Gemini API response."""
|
59 |
+
try:
|
60 |
+
return response["candidates"][0]["content"]["parts"][0]["text"]
|
61 |
+
except (KeyError, IndexError):
|
62 |
+
return "No analysis available"
|
63 |
+
|
64 |
+
# --------------------------
|
65 |
+
# Helper Functions
|
66 |
+
# --------------------------
|
67 |
+
def frame_to_base64(frame):
|
68 |
+
"""Convert an image frame to base64 format."""
|
69 |
+
_, buffer = cv2.imencode(".jpg", frame)
|
70 |
+
return base64.b64encode(buffer).decode("utf-8")
|
71 |
+
|
72 |
+
def extract_video_frame(video_path, timestamp):
|
73 |
+
"""Extract a frame at a specific timestamp from the video."""
|
74 |
+
cap = cv2.VideoCapture(video_path)
|
75 |
+
cap.set(cv2.CAP_PROP_POS_MSEC, timestamp * 1000)
|
76 |
+
success, frame = cap.read()
|
77 |
+
cap.release()
|
78 |
+
return frame if success else None
|
79 |
+
|
80 |
+
def download_video(video_url):
|
81 |
+
"""Download video from URL to a temporary file."""
|
82 |
+
try:
|
83 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as temp_file:
|
84 |
+
response = requests.get(video_url, stream=True, timeout=30)
|
85 |
+
response.raise_for_status()
|
86 |
+
for chunk in response.iter_content(chunk_size=8192):
|
87 |
+
temp_file.write(chunk)
|
88 |
+
return temp_file.name
|
89 |
+
except Exception as e:
|
90 |
+
st.error(f"Video download failed: {str(e)}")
|
91 |
+
return None
|
92 |
+
|
93 |
+
# --------------------------
|
94 |
+
# Streamlit App
|
95 |
+
# --------------------------
|
96 |
+
st.title("🎥 Military Surveillance AI")
|
97 |
+
st.write("Upload a battlefield surveillance video to analyze.")
|
98 |
+
|
99 |
+
video_url = st.text_input("Enter Video URL:")
|
100 |
+
if st.button("Analyze Video") and video_url:
|
101 |
+
st.info("Downloading video...")
|
102 |
+
video_path = download_video(video_url)
|
103 |
+
|
104 |
+
if video_path:
|
105 |
+
client = GeminiClient(API_KEY, API_URL)
|
106 |
+
st.success("Video downloaded successfully!")
|
107 |
+
|
108 |
+
st.info("Processing video and analyzing frames...")
|
109 |
+
log = []
|
110 |
+
|
111 |
+
# Extract frames and analyze
|
112 |
+
for timestamp in range(10, 40, 10): # Analyze at 10s, 20s, 30s
|
113 |
+
frame = extract_video_frame(video_path, timestamp)
|
114 |
+
if frame is None:
|
115 |
+
log.append(f"[{timestamp}s] Frame extraction failed ❌")
|
116 |
+
continue
|
117 |
+
|
118 |
+
analysis = client.analyze_frame(frame_to_base64(frame), timestamp)
|
119 |
+
log.append(f"[{timestamp}s] {analysis}")
|
120 |
+
st.write(f"### Timestamp: {timestamp}s")
|
121 |
+
st.image(frame, caption=f"Frame at {timestamp}s", use_column_width=True)
|
122 |
+
st.write(analysis)
|
123 |
+
time.sleep(2) # Simulate processing delay
|
124 |
+
|
125 |
+
st.success("Analysis complete! ✅")
|
126 |
+
st.text_area("Summary of Analysis:", "\n".join(log), height=200)
|
127 |
+
|