Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -7,13 +7,6 @@ from moviepy.editor import VideoFileClip
|
|
7 |
import json
|
8 |
import spaces
|
9 |
import torch
|
10 |
-
import random
|
11 |
-
import string
|
12 |
-
from datetime import datetime, timedelta
|
13 |
-
from collections import defaultdict
|
14 |
-
|
15 |
-
# Set up logging
|
16 |
-
logging.basicConfig(level=logging.INFO, filename="api_access.log", format="%(asctime)s - %(message)s")
|
17 |
|
18 |
torch.set_num_threads(1)
|
19 |
torch.set_num_interop_threads(1)
|
@@ -23,95 +16,19 @@ torch.backends.cudnn.benchmark = False
|
|
23 |
torch.backends.cuda.matmul.allow_tf32 = False
|
24 |
torch.backends.cudnn.allow_tf32 = False
|
25 |
|
26 |
-
|
27 |
-
ip_access_records = defaultdict(lambda: {
|
28 |
-
'total_access_time': timedelta(),
|
29 |
-
'blocked_until': None,
|
30 |
-
'session_start': None,
|
31 |
-
'last_access_time': None
|
32 |
-
})
|
33 |
|
34 |
-
def truncate_video(video_file):
|
35 |
-
"""Truncates video to 30 seconds and saves it as a temporary file."""
|
36 |
-
clip = VideoFileClip(video_file)
|
37 |
-
truncated_clip = clip.subclip(0, min(30, clip.duration)) # Change 15 to 30 for 30 seconds
|
38 |
-
truncated_video_file = "temp_truncated_video.mp4"
|
39 |
-
truncated_clip.write_videofile(truncated_video_file, codec="libx264", audio_codec="aac")
|
40 |
-
clip.close()
|
41 |
-
truncated_clip.close()
|
42 |
-
return truncated_video_file
|
43 |
-
|
44 |
-
'''
|
45 |
-
AUTHORIZED_API_KEY = os.getenv("HF_API_KEY")
|
46 |
-
|
47 |
-
def is_authorized(request: gr.Request):
|
48 |
-
"""Checks if the request includes the correct API key."""
|
49 |
-
api_key = request.headers.get("API-Key")
|
50 |
-
print("API-Key received:", api_key) # Log the received API key for debugging
|
51 |
-
return api_key == AUTHORIZED_API_KEY
|
52 |
-
'''
|
53 |
-
|
54 |
-
SESSION_TIMEOUT = timedelta(hours=1)
|
55 |
-
|
56 |
-
# Function to log API access attempts
|
57 |
-
def log_api_access(successful_captcha, video_file=None, client_ip="unknown", blocked=False):
|
58 |
-
status = "Blocked" if blocked else "Allowed"
|
59 |
-
logging.info(f"Access Attempt - Status: {status}, Successful CAPTCHA: {successful_captcha}, "
|
60 |
-
f"Video File: {video_file if video_file else 'N/A'}, Client IP: {client_ip}")
|
61 |
-
|
62 |
-
# Function to get client IP address
|
63 |
-
def get_client_ip(request: gr.Request):
|
64 |
-
return request.client.host
|
65 |
-
|
66 |
-
# Function to check if IP is blocked
|
67 |
-
def is_ip_blocked(client_ip):
|
68 |
-
record = ip_access_records[client_ip]
|
69 |
-
blocked_until = record.get('blocked_until')
|
70 |
-
if blocked_until and datetime.now() < blocked_until:
|
71 |
-
return True
|
72 |
-
return False
|
73 |
-
|
74 |
-
# Function to update IP access time
|
75 |
-
def update_ip_access_time(client_ip):
|
76 |
-
record = ip_access_records[client_ip]
|
77 |
-
now = datetime.now()
|
78 |
-
session_start = record.get('session_start')
|
79 |
-
|
80 |
-
if session_start:
|
81 |
-
elapsed = now - session_start
|
82 |
-
record['total_access_time'] += elapsed
|
83 |
-
record['session_start'] = now
|
84 |
-
else:
|
85 |
-
record['session_start'] = now
|
86 |
-
|
87 |
-
# Remove access times older than 24 hours
|
88 |
-
last_access_time = record.get('last_access_time')
|
89 |
-
if last_access_time and now - last_access_time > timedelta(hours=24):
|
90 |
-
record['total_access_time'] = timedelta()
|
91 |
-
|
92 |
-
record['last_access_time'] = now
|
93 |
-
|
94 |
-
# Check if total access time in last 24 hours exceeds 1 hour
|
95 |
-
if record['total_access_time'] >= timedelta(hours=1):
|
96 |
-
# Block IP for 48 hours
|
97 |
-
record['blocked_until'] = now + timedelta(hours=48)
|
98 |
-
record['total_access_time'] = timedelta() # Reset total access time
|
99 |
-
|
100 |
-
# Function to truncate video
|
101 |
def truncate_video(video_file):
|
102 |
"""Truncates video to 15 seconds and saves it as a temporary file."""
|
103 |
clip = VideoFileClip(video_file)
|
104 |
truncated_clip = clip.subclip(0, min(15, clip.duration))
|
105 |
truncated_video_file = "temp_truncated_video.mp4"
|
106 |
truncated_clip.write_videofile(truncated_video_file, codec="libx264", audio_codec="aac")
|
107 |
-
clip.close()
|
108 |
-
truncated_clip.close()
|
109 |
return truncated_video_file
|
110 |
|
111 |
-
# Clone repository
|
112 |
def clone_repo():
|
113 |
"""Clone the GitHub repository containing the backend."""
|
114 |
-
repo_url = "https://github.com/NeeravSood/AllMark-MVP.git"
|
115 |
repo_path = "./repository"
|
116 |
|
117 |
github_pat = os.getenv("GITHUB_PAT")
|
@@ -135,14 +52,13 @@ def clone_repo():
|
|
135 |
print("Error:", e.stderr)
|
136 |
raise RuntimeError(f"Failed to clone repository: {e.stderr}")
|
137 |
|
138 |
-
# Import backend script
|
139 |
def import_backend_script(script_name):
|
140 |
"""Dynamically import the backend script."""
|
141 |
try:
|
142 |
script_path = os.path.join("./repository", script_name)
|
143 |
if not os.path.exists(script_path):
|
144 |
raise FileNotFoundError(f"Script {script_name} not found in the repository.")
|
145 |
-
|
146 |
spec = importlib.util.spec_from_file_location("backend_module", script_path)
|
147 |
backend_module = importlib.util.module_from_spec(spec)
|
148 |
spec.loader.exec_module(backend_module)
|
@@ -151,63 +67,31 @@ def import_backend_script(script_name):
|
|
151 |
logging.error(f"Error importing backend script: {str(e)}")
|
152 |
raise RuntimeError(f"Failed to import backend script: {str(e)}")
|
153 |
|
154 |
-
# Clone and import repository
|
155 |
clone_repo()
|
156 |
-
backend = import_backend_script("app.py")
|
157 |
analyzer = backend.DeepfakeAnalyzer()
|
158 |
|
159 |
-
# Generate a random CAPTCHA challenge
|
160 |
-
def generate_captcha():
|
161 |
-
return ''.join(random.choices(string.ascii_uppercase + string.digits, k=5))
|
162 |
-
|
163 |
-
# Function to verify CAPTCHA per user session
|
164 |
-
def verify_captcha(user_input, captcha_solution):
|
165 |
-
if user_input.strip() == captcha_solution:
|
166 |
-
return "Captcha verified. Video analysis will proceed.", True
|
167 |
-
else:
|
168 |
-
return "Incorrect CAPTCHA. Please try again.", False
|
169 |
-
|
170 |
@spaces.GPU(duration=1000)
|
171 |
-
def analyze_video(video_file
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
client_ip = get_client_ip(request)
|
177 |
-
|
178 |
-
# Check if IP is blocked
|
179 |
-
if is_ip_blocked(client_ip):
|
180 |
-
log_api_access(successful_captcha=False, video_file=os.path.basename(video_file) if video_file else "N/A", client_ip=client_ip, blocked=True)
|
181 |
-
return {"error": "Your IP has been blocked due to excessive usage. Please try again later."}
|
182 |
-
|
183 |
-
# Verify CAPTCHA
|
184 |
-
message, success = verify_captcha(captcha_input, captcha_solution)
|
185 |
-
if not success:
|
186 |
-
log_api_access(successful_captcha=False, video_file=os.path.basename(video_file) if video_file else "N/A", client_ip=client_ip)
|
187 |
-
return {"error": message}
|
188 |
-
|
189 |
-
# Update IP access time
|
190 |
-
update_ip_access_time(client_ip)
|
191 |
-
|
192 |
-
# Log API access attempt with successful CAPTCHA
|
193 |
-
log_api_access(successful_captcha=True, video_file=os.path.basename(video_file) if video_file else "N/A", client_ip=client_ip)
|
194 |
|
195 |
try:
|
196 |
-
# Video truncation and analysis
|
197 |
truncated_video = truncate_video(video_file)
|
198 |
results = analyzer.analyze_media(truncated_video)
|
199 |
|
200 |
-
#
|
201 |
combined_assessment = results.get('combined_assessment', 0)
|
202 |
if isinstance(combined_assessment, str) and combined_assessment.lower() == "deepfake":
|
203 |
analysis_result = "a deepfake"
|
204 |
else:
|
205 |
combined_assessment = int(combined_assessment) if str(combined_assessment).isdigit() else 0
|
206 |
analysis_result = "genuine/original" if combined_assessment < 50 else "a deepfake"
|
207 |
-
|
208 |
output = {
|
209 |
"message": f"According to our analysis, the video you uploaded appears to be {analysis_result}. "
|
210 |
-
f"
|
211 |
}
|
212 |
return output
|
213 |
|
@@ -215,59 +99,14 @@ def analyze_video(video_file, captcha_input, captcha_solution, request: gr.Reque
|
|
215 |
logging.error(f"Error during analysis: {e}")
|
216 |
return {"error": "An error occurred during video analysis. Please check your input and try again."}
|
217 |
|
218 |
-
# Interface with CAPTCHA after video upload
|
219 |
-
def main_interface():
|
220 |
-
with gr.Blocks() as interface:
|
221 |
-
gr.Markdown("# AllMark - Deepfake Analyzer")
|
222 |
-
gr.Markdown("Upload a video to proceed with analysis.")
|
223 |
-
|
224 |
-
# State variables for per-session data
|
225 |
-
captcha_solution = gr.State()
|
226 |
-
|
227 |
-
# Video input with display size restriction
|
228 |
-
video_input = gr.Video(label="Upload Video", height=300) # Adjust height as needed
|
229 |
-
captcha_text = gr.Textbox(label="CAPTCHA", interactive=False, visible=False)
|
230 |
-
captcha_input = gr.Textbox(label="Enter CAPTCHA Here", visible=False)
|
231 |
-
captcha_output = gr.Textbox(label="CAPTCHA Status", interactive=False, visible=False)
|
232 |
-
analyze_button = gr.Button("Analyze Video", visible=False)
|
233 |
-
analysis_output = gr.JSON(label="Analysis Result")
|
234 |
-
|
235 |
-
# Function to show CAPTCHA after video upload
|
236 |
-
def show_captcha(video):
|
237 |
-
if video is not None:
|
238 |
-
# Generate a new CAPTCHA for this session
|
239 |
-
new_captcha = generate_captcha()
|
240 |
-
return (
|
241 |
-
gr.update(visible=True, value=new_captcha),
|
242 |
-
gr.update(visible=True),
|
243 |
-
gr.update(visible=True),
|
244 |
-
gr.update(visible=True),
|
245 |
-
new_captcha # Update the session state
|
246 |
-
)
|
247 |
-
else:
|
248 |
-
return (
|
249 |
-
gr.update(visible=False),
|
250 |
-
gr.update(visible=False),
|
251 |
-
gr.update(visible=False),
|
252 |
-
gr.update(visible=False),
|
253 |
-
gr.no_update
|
254 |
-
)
|
255 |
-
|
256 |
-
video_input.change(
|
257 |
-
show_captcha,
|
258 |
-
inputs=video_input,
|
259 |
-
outputs=[captcha_text, captcha_input, captcha_output, analyze_button, captcha_solution]
|
260 |
-
)
|
261 |
-
|
262 |
-
# Handle analysis
|
263 |
-
analyze_button.click(
|
264 |
-
analyze_video,
|
265 |
-
inputs=[video_input, captcha_input, captcha_solution],
|
266 |
-
outputs=analysis_output
|
267 |
-
)
|
268 |
|
269 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
270 |
|
271 |
-
# Launch interface
|
272 |
if __name__ == "__main__":
|
273 |
-
|
|
|
7 |
import json
|
8 |
import spaces
|
9 |
import torch
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
torch.set_num_threads(1)
|
12 |
torch.set_num_interop_threads(1)
|
|
|
16 |
torch.backends.cuda.matmul.allow_tf32 = False
|
17 |
torch.backends.cudnn.allow_tf32 = False
|
18 |
|
19 |
+
ACCESS_KEY = os.getenv("ACCESS_KEY")
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
def truncate_video(video_file):
|
22 |
"""Truncates video to 15 seconds and saves it as a temporary file."""
|
23 |
clip = VideoFileClip(video_file)
|
24 |
truncated_clip = clip.subclip(0, min(15, clip.duration))
|
25 |
truncated_video_file = "temp_truncated_video.mp4"
|
26 |
truncated_clip.write_videofile(truncated_video_file, codec="libx264", audio_codec="aac")
|
|
|
|
|
27 |
return truncated_video_file
|
28 |
|
|
|
29 |
def clone_repo():
|
30 |
"""Clone the GitHub repository containing the backend."""
|
31 |
+
repo_url = "https://github.com/NeeravSood/AllMark-MVP.git" # Update when changing
|
32 |
repo_path = "./repository"
|
33 |
|
34 |
github_pat = os.getenv("GITHUB_PAT")
|
|
|
52 |
print("Error:", e.stderr)
|
53 |
raise RuntimeError(f"Failed to clone repository: {e.stderr}")
|
54 |
|
|
|
55 |
def import_backend_script(script_name):
|
56 |
"""Dynamically import the backend script."""
|
57 |
try:
|
58 |
script_path = os.path.join("./repository", script_name)
|
59 |
if not os.path.exists(script_path):
|
60 |
raise FileNotFoundError(f"Script {script_name} not found in the repository.")
|
61 |
+
|
62 |
spec = importlib.util.spec_from_file_location("backend_module", script_path)
|
63 |
backend_module = importlib.util.module_from_spec(spec)
|
64 |
spec.loader.exec_module(backend_module)
|
|
|
67 |
logging.error(f"Error importing backend script: {str(e)}")
|
68 |
raise RuntimeError(f"Failed to import backend script: {str(e)}")
|
69 |
|
|
|
70 |
clone_repo()
|
71 |
+
backend = import_backend_script("app.py")
|
72 |
analyzer = backend.DeepfakeAnalyzer()
|
73 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
74 |
@spaces.GPU(duration=1000)
|
75 |
+
def analyze_video(video_file):
|
76 |
+
if access_key != ACCESS_KEY:
|
77 |
+
logging.error("Invalid access key provided.")
|
78 |
+
return {"error": "Unauthorized access. Invalid access key."}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
79 |
|
80 |
try:
|
|
|
81 |
truncated_video = truncate_video(video_file)
|
82 |
results = analyzer.analyze_media(truncated_video)
|
83 |
|
84 |
+
# Get combined_assessment and handle non-numeric values
|
85 |
combined_assessment = results.get('combined_assessment', 0)
|
86 |
if isinstance(combined_assessment, str) and combined_assessment.lower() == "deepfake":
|
87 |
analysis_result = "a deepfake"
|
88 |
else:
|
89 |
combined_assessment = int(combined_assessment) if str(combined_assessment).isdigit() else 0
|
90 |
analysis_result = "genuine/original" if combined_assessment < 50 else "a deepfake"
|
91 |
+
|
92 |
output = {
|
93 |
"message": f"According to our analysis, the video you uploaded appears to be {analysis_result}. "
|
94 |
+
f"{len(results['video_analysis']['frame_results'])} frames were analyzed in total."
|
95 |
}
|
96 |
return output
|
97 |
|
|
|
99 |
logging.error(f"Error during analysis: {e}")
|
100 |
return {"error": "An error occurred during video analysis. Please check your input and try again."}
|
101 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
102 |
|
103 |
+
interface = gr.Interface(
|
104 |
+
fn=analyze_video,
|
105 |
+
inputs=gr.Video(label="Upload Video"),
|
106 |
+
outputs="json",
|
107 |
+
title="AllMark - Deepfake Analyzer",
|
108 |
+
description="Upload a video to analyze. N.B. - Only mp4 files. Processing time 1-10 minutes. For any false negatives, please contact the publisher for verification. Incognito Mode Recommended"
|
109 |
+
)
|
110 |
|
|
|
111 |
if __name__ == "__main__":
|
112 |
+
interface.launch()
|