update-logic

.gitignore

@@ -1,3 +1,2 @@
 __pycache__/
-/users/*
-!/users/face_features.csv
+/users/*

faceVerificationUtilization/setConfig.py

@@ -44,6 +44,4 @@ transform = transforms.Compose([
     transforms.ToTensor(),
     transforms.Normalize([0.485, 0.456, 0.406], 
                          [0.229, 0.224, 0.225])  
-])
-
-faiss_index, labels, db = load_db("users/face_features.csv")
+])

faceVerificationUtilization/utils.py

@@ -3,7 +3,7 @@ import numpy as np
 from PIL import Image
 import torch
 import pandas as pd
-from .setConfig import efficientnet_model, face_detector, transform, pca_xgb, faiss, faiss_index, labels, db
+from .setConfig import efficientnet_model, face_detector, transform, pca_xgb, faiss, load_db
 import os
 
 def ImgPreprocessing(img):
@@ -17,7 +17,8 @@ def ImgPreprocessing(img):
     return img
 
 def YoloFaceDetection(img):
-    results = face_detector.predict(img, conf=0.8)
+    results = face_detector.predict(img, conf=0.7)
+
     keyReturn = {
         'message': "",
         'status': False,
@@ -28,8 +29,11 @@ def YoloFaceDetection(img):
         keyReturn['message'] = "Tidak ada wajah terdeteksi"
         return keyReturn
 
-    box = results[0].boxes[0]
-    x1, y1, x2, y2 = map(int, box.xyxy[0].cpu().numpy())
+    boxes = results[0].boxes.xyxy.cpu().numpy()
+    confs = results[0].boxes.conf.cpu().numpy()
+
+    max_idx = confs.argmax()
+    x1, y1, x2, y2 = map(int, boxes[max_idx])
 
     keyReturn['message'] = "Face detected"
     keyReturn['status'] = True
@@ -41,9 +45,10 @@ def FaceValidationPredict(**face_crop):
     x1, y1, x2, y2 = face_crop['coordinate']
     face_crop = face_crop['img'][y1:y2, x1:x2]
 
-    face_crop = cv.cvtColor(face_crop, cv.COLOR_BGR2RGB)  
-    
+    face_crop = cv.cvtColor(face_crop, cv.COLOR_BGR2GRAY)
+    face_crop = cv.cvtColor(face_crop, cv.COLOR_GRAY2RGB)  
     face_pil = Image.fromarray(face_crop)
+
     face_tensor = transform(face_pil).unsqueeze(0)
 
     with torch.no_grad():
@@ -51,7 +56,7 @@ def FaceValidationPredict(**face_crop):
 
     pred = pca_xgb.predict(features)[0]
 
-    return pred, features
+    return pred, features, face_crop
 
 def FaceValidation(frame: np.ndarray):
     if frame is None:
@@ -70,7 +75,7 @@ def FaceValidation(frame: np.ndarray):
     results = keyReturn['coordinate']
     x1, y1, x2, y2 = results
 
-    validation, features = FaceValidationPredict(coordinate=[x1, y1, x2, y2], img=img)
+    validation, features, face_crop = FaceValidationPredict(coordinate=[x1, y1, x2, y2], img=img)
     pred = 'Wajah Valid' if validation == 0 else 'Wajah Tidak Valid'
 
     return f"Predicted class: {pred}"
@@ -90,13 +95,13 @@ def FaceRecord(img, name, photo_idx):
         return keyReturn['message']
 
     x1, y1, x2, y2 = keyReturn['coordinate']
-    pred, features = FaceValidationPredict(coordinate=[x1, y1, x2, y2], img=img)
+    pred, features, face_crop = FaceValidationPredict(coordinate=[x1, y1, x2, y2], img=img)
 
     if pred == 1:  
         return f"[Foto {photo_idx}] Gagal: wajah tidak valid"
 
     save_path = os.path.join(user_dir, f"photo_{photo_idx}.jpg")
-    cv.imwrite(save_path, img)
+    cv.imwrite(save_path, face_crop)
 
     csv_path = "users/face_features.csv"
     row = pd.DataFrame({
@@ -130,12 +135,14 @@ def Recognize(frame: np.ndarray):
 
     results = keyReturn['coordinate']
     x1, y1, x2, y2 = results
-    pred, features = FaceValidationPredict(coordinate=[x1, y1, x2, y2], img=img)
+    pred, features, face_crop = FaceValidationPredict(coordinate=[x1, y1, x2, y2], img=img)
     faiss.normalize_L2(features)
 
     if pred == 1:
         return 'Wajah tidak terdeteksi'
     
+    faiss_index, labels, db = load_db("users/face_features.csv")
+
     if faiss_index is None:
         return "Database kosong"
 
@@ -143,7 +150,7 @@ def Recognize(frame: np.ndarray):
     score = float(D[0][0])
     idx = int(I[0][0])
 
-    if score < 0.7:  
+    if score < 0.65:  
         return f"Tidak dikenali"
     else:
         return f"Terkenali sebagai: {labels[idx]} - (score={score:.2f})"