Update app.py

app.py

@@ -1,53 +1,47 @@
 import os
-from transformers import AutoModel, AutoTokenizer
-import torch
-from PIL import Image
-
-# Load pre-trained model and tokenizer
-model_name = "deepface/face-recognition"
-model = AutoModel.from_pretrained(model_name)
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-
-# Define folders
-folder1 = "C:/Users/nikit/OneDrive/esktop/New folder (2)"
-folder2 = "C:/Users/nikit/OneDrive/Desktop/New folder"
-
-# Define threshold for matching (e.g., 0.5)
-threshold = 0.5
-
-# Function to extract face embeddings
-def extract_embeddings(folder):
-    embeddings = []
-    for file in os.listdir(folder):
-        image = Image.open(os.path.join(folder, file))
-        inputs = tokenizer(images=image, return_tensors='pt')
-        outputs = model(**inputs)
-        embeddings.append(outputs.last_hidden_state[:, 0, :].detach().numpy())
-    return embeddings
-
-# Extract embeddings from both folders
-embeddings1 = extract_embeddings(folder1)
-embeddings2 = extract_embeddings(folder2)
-
-# Compare embeddings
-matches = []
-for embedding1 in embeddings1:
-    for embedding2 in embeddings2:
-        similarity = torch.cosine_similarity(torch.tensor(embedding1), torch.tensor(embedding2))
-        if similarity > threshold:
-            matches.append((embedding1, embedding2))
-
-# Print matches
-for match in matches:
-    print(f"Match found: {match[0]} (Folder 1) and {match[1]} (Folder 2)")
-
-# Print non-matches
-non_matches = []
-for embedding1 in embeddings1:
-    for embedding2 in embeddings2:
-        similarity = torch.cosine_similarity(torch.tensor(embedding1), torch.tensor(embedding2))
-        if similarity < threshold:
-            non_matches.append((embedding1, embedding2))
-
-for non_match in non_matches:
-    print(f"No match found: {non_match[0]} (Folder 1) and {non_match[1]} (Folder 2)")
+import cv2
+import numpy as np
+
+def compare_images(image1, image2):
+    # Read the images
+    img1 = cv2.imread(r"C:\Users\nikit\OneDrive\Desktop\Test1")
+    img2 = cv2.imread(r"C:\Users\nikit\OneDrive\Desktop\Test2")
+
+    # Check if both images are loaded properly
+    if img1 is None or img2 is None:
+        return False
+
+    # Resize images to the same size
+    img1 = cv2.resize(img1, (100, 100))
+    img2 = cv2.resize(img2, (100, 100))
+
+    # Convert images to grayscale
+    gray1 = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
+    gray2 = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
+
+    # Calculate the Structural Similarity Index (SSI)
+    score, _ = cv2.compare_ssim(gray1, gray2, full=True)
+    
+    # You can adjust the threshold based on your needs
+    return score >= 0.9  # Threshold for similarity
+
+def compare_folders(staff_folder, unknown_folder):
+    for staff_image in os.listdir(staff_folder):
+        staff_image_path = os.path.join(staff_folder, staff_image)
+        found_match = False
+        
+        for unknown_image in os.listdir(unknown_folder):
+            unknown_image_path = os.path.join(unknown_folder, unknown_image)
+            
+            if compare_images(staff_image_path, unknown_image_path):
+                print(f"Match found: {staff_image} matches {unknown_image} (Staff)")
+                found_match = True
+                break
+        
+        if not found_match:
+            print(f"No match found for {staff_image} (Unknown)")
+
+# Usage
+staff_folder_path = 'path/to/staff/images'
+unknown_folder_path = 'path/to/unknown/images'
+compare_folders(staff_folder_path, unknown_folder_path)