Update app.py

app.py

@@ -1,41 +1,58 @@
 import gradio as gr
-import numpy as np
 import cv2
-from sklearn.metrics.pairwise import euclidean_distances
-
-# Load model files
-mean_face = np.load("model/mean_face.npy")
-eigenfaces = np.load("model/eigenfaces.npy")
-projections = np.load("model/projections.npy")
-labels = np.load("model/labels.npy")
-label_names = np.load("model/label_names.npy")
-with open("model/img_shape.txt", "r") as f:
-    img_h, img_w = map(int, f.read().strip().split())
-
-THRESHOLD = 8000
-
-def recognize_face_from_frame(frame):
-    img = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
-    img = cv2.resize(img, (img_w, img_h))
-    img_flat = img.flatten()
-    img_centered = img_flat - mean_face
-    projected = np.dot(img_centered, eigenfaces.T)
-    dists = euclidean_distances([projected], projections)[0]
-    min_dist = np.min(dists)
-    min_index = np.argmin(dists)
-
-    if min_dist < THRESHOLD:
-        predicted_name = label_names[labels[min_index]]
-        return f"🧠 Predicted: {predicted_name} (distance: {min_dist:.2f})"
-    else:
-        return "😕 Unknown face"
+import numpy as np
+import os
+
+# Load your trained model and database here
+# Example: load mean face, eigenfaces, and face projections
+mean = np.load("mean.npy")
+eigenfaces = np.load("eigenfaces.npy")
+projected_faces = np.load("projected_faces.npy")
+labels = np.load("labels.npy")  # Assuming you saved labels for your dataset
+
+def preprocess_image(image):
+    # Convert to grayscale
+    gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
+    resized = cv2.resize(gray, (100, 100)).flatten()
+    normalized = resized.astype('float32')
+    return normalized
+
+def project_face(face_vector, mean, eigenfaces):
+    centered_vector = face_vector - mean
+    return np.dot(eigenfaces, centered_vector)
+
+def recognize_face(uploaded_image):
+    try:
+        if uploaded_image is None:
+            return "No image provided."
+
+        face_vector = preprocess_image(uploaded_image)
+        projected = project_face(face_vector, mean, eigenfaces)
+
+        # Compare with database
+        min_dist = float('inf')
+        matched_label = "Unknown"
+        for idx, known_proj in enumerate(projected_faces):
+            dist = np.linalg.norm(projected - known_proj)
+            if dist < min_dist:
+                min_dist = dist
+                matched_label = labels[idx]
+
+        threshold = 3000  # Adjust based on your model
+        if min_dist < threshold:
+            return f"Matched: {matched_label} (Distance: {min_dist:.2f})"
+        else:
+            return "No match found"
+    except Exception as e:
+        return f"Error: {str(e)}"
 
 iface = gr.Interface(
-    fn=recognize_face_from_frame,
-    inputs=gr.Image(type="numpy"),
-    outputs="text",
-    title="🔍 EigenFace Recognition",
-    description="Upload a face image (webcam capture available on some devices)."
+    fn=recognize_face,
+    inputs=gr.Image(type="numpy", label="Take or upload a photo"),
+    outputs=gr.Textbox(label="Prediction"),
+    title="Eigenface Face Recognition",
+    description="Upload or take a photo. The model will identify the person if in the database.",
+    allow_flagging="never",
 )
 
 iface.launch()