Added temp file for audio and other folders for faces etc

app.py

@@ -15,6 +15,7 @@ from pytorch_grad_cam.utils.model_targets import ClassifierOutputTarget
 from pytorch_grad_cam.utils.image import show_cam_on_image
 import os
 import warnings
+import tempfile
 import glob
 from concurrent.futures import ThreadPoolExecutor
 import multiprocessing
@@ -72,7 +73,7 @@ def audiopredict(audio):
     if audio is not None:
         audio_clip = load_audio(audio)
         ai_generated_probability = classify_audio_clip(audio_clip)
-        image_path = os.path.join("D:\\Dhejan\\HACKATHONS\\Cyber hackathon - SRM RMP\\Projects\\sources\\WhatsApp Image 2024-08-14 at 10.39.08_2283f2a8.jpg")
+        image_path = os.path.join("./wave.jpg")
         image = Image.open(image_path)
         if ai_generated_probability < 0.5:
             return "Real", "The audio is likely to be Real", "No EXIF data found in the audio", image
@@ -81,43 +82,41 @@ def audiopredict(audio):
     
 # Video Input Code
 def save_video(video_path):
-    
-    # Modify this path to your desired folder
-    save_folder = "D:\\Dhejan\\HACKATHONS\\Cyber hackathon - SRM RMP\\Projects\\video"
-
-    # Extract filename from path
-    filename = video_path.split("\\")[-1]
-
-    # Save video to specified folder
-    with open(f"{save_folder}/{filename}", "wb") as f:
-        f.write(open(video_path, "rb").read())
-
-    # Process frames, select faces, and perform deepfake identification
-    textoutput, exif, face_with_mask = process_video(save_folder, filename)   
-    print(textoutput)
-    string = textoutput
-
-    # Extract percentages and convert them to floats
-    percentages = re.findall(r"(\d+\.\d+)%", string)
-    real_percentage = float(percentages[0])
-    fake_percentage = float(percentages[1])
-
-    # Determine which percentage is higher
-    if real_percentage > fake_percentage:
-        print("Real")
-        val = "Real"
-    else:
-        print("Fake")
-        val = "Deepfake"
+    # Create a temporary directory to save the video
+    with tempfile.TemporaryDirectory() as temp_dir:
+        # Extract filename from path
+        filename = os.path.basename(video_path)
+
+        # Save video to the temporary folder
+        temp_video_path = os.path.join(temp_dir, filename)
+        with open(temp_video_path, "wb") as f:
+            f.write(open(video_path, "rb").read())
+
+        # Process frames, select faces, and perform deepfake identification
+        textoutput, exif, face_with_mask = process_video(temp_dir, filename)   
+        print(textoutput)
+        string = textoutput
+
+        # Extract percentages and convert them to floats
+        percentages = re.findall(r"(\d+\.\d+)%", string)
+        real_percentage = float(percentages[0])
+        fake_percentage = float(percentages[1])
+
+        # Determine which percentage is higher
+        if real_percentage > fake_percentage:
+            print("Real")
+            val = "Real"
+        else:
+            print("Fake")
+            val = "Deepfake"
 
-    
     return val, textoutput, exif, face_with_mask
 
 def process_video(video_folder, video_filename):
     # Additional Processing (Frames, Faces, Deepfake Identification)
-    frames_base_dir = "D:\\Dhejan\\HACKATHONS\\Cyber hackathon - SRM RMP\\Projects\\frames"
-    faces_base_dir = "D:\\Dhejan\\HACKATHONS\\Cyber hackathon - SRM RMP\\Projects\\faces"
-    selected_faces_base_dir = "D:\\Dhejan\\HACKATHONS\\Cyber hackathon - SRM RMP\\Projects\\selected_faces"
+    frames_base_dir = "./frames"
+    faces_base_dir = "./faces"
+    selected_faces_base_dir = "./selected_faces"
 
     # Find the latest video
     video_path = os.path.join(video_folder, video_filename)
@@ -213,7 +212,7 @@ def identify_deepfake(selected_faces_dir):
     model = InceptionResnetV1(pretrained="vggface2", classify=True, num_classes=1, device=DEVICE)
 
     # Load the model checkpoint
-    checkpoint_path = "D:\\Dhejan\\HACKATHONS\\Cyber hackathon - SRM RMP\\Projects\\resnetinceptionv1_epoch_32.pth"  # Update this path
+    checkpoint_path = "./resnetinceptionv1_epoch_32.pth"  # Update this path
     checkpoint = torch.load(checkpoint_path, map_location=DEVICE)
     model.load_state_dict(checkpoint['model_state_dict'])
     model.to(DEVICE)
@@ -325,7 +324,7 @@ def predictimage(input_image: Image.Image):
         device=DEVICE
     )
 
-    checkpoint = torch.load("D:\\Dhejan\\HACKATHONS\\Cyber hackathon - SRM RMP\\Projects\\resnetinceptionv1_epoch_32.pth", map_location=torch.device('cpu'))
+    checkpoint = torch.load("./resnetinceptionv1_epoch_32.pth", map_location=torch.device('cpu'))
     model.load_state_dict(checkpoint['model_state_dict'])
     model.to(DEVICE)
     model.eval()