'try-to-test'

app.py

@@ -0,0 +1,4 @@
+from faceVerificationUtilization import demo
+
+if __name__ == "__main__":
+    demo.launch()

faceVerificationModel/efficientnetv2_s_features.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:431d1c7064464949b5895965e09db9c1c265bb0c14524804ac626439546a87d9
+size 81622458

faceVerificationModel/pca_xgb_pipeline.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5c1181564c00e26d980cf8d30770fd8a678e3cb396327e6fa68c83cfef75b3b
+size 357574

faceVerificationUtilization/__init__.py

@@ -0,0 +1 @@
+from .utils import demo

faceVerificationUtilization/setConfig.py

@@ -0,0 +1,25 @@
+from ultralytics import YOLO
+import torch, joblib
+from huggingface_hub import hf_hub_download
+from torchvision import transforms
+import torchvision.models as models
+import torch.nn as nn
+
+model_path = hf_hub_download(repo_id="arnabdhar/YOLOv8-Face-Detection", filename="model.pt")
+face_detector = YOLO(model_path)
+
+efficientnet_model = models.efficientnet_v2_s(weights=None)
+efficientnet_model.classifier = nn.Identity()
+
+state_dict = torch.load("faceVerificationModel/efficientnetv2_s_features.pth", map_location="cpu")
+efficientnet_model.load_state_dict(state_dict)
+efficientnet_model.eval()
+
+pca_xgb = joblib.load("faceVerificationModel/pca_xgb_pipeline.pkl")
+
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),   
+    transforms.ToTensor(),
+    transforms.Normalize([0.485, 0.456, 0.406], 
+                         [0.229, 0.224, 0.225])  
+])

faceVerificationUtilization/utils.py

@@ -0,0 +1,52 @@
+import cv2 as cv
+import numpy as np
+from PIL import Image
+import torch
+from .setConfig import efficientnet_model, face_detector, transform, pca_xgb
+import gradio as gr
+
+def ImgPreprocessing(img):
+    if len(img.shape) == 2 or img.shape[2] == 1:
+        img = cv.cvtColor(img, cv.COLOR_GRAY2BGR)
+
+    img_yuv = cv.cvtColor(img, cv.COLOR_BGR2YUV)
+    clahe = cv.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
+    img_yuv[:, :, 0] = clahe.apply(img_yuv[:, :, 0])
+    img = cv.cvtColor(img_yuv, cv.COLOR_YUV2BGR)
+    return img
+
+def predict(frame: np.ndarray):
+    if frame is None:
+        return "No frame captured from webcam"
+    
+    if isinstance(frame, dict):
+        frame = frame['image']
+
+    img = ImgPreprocessing(frame)
+
+    results = face_detector.predict(img)
+    if len(results[0].boxes) == 0:
+        return "No face detected"
+
+    x1, y1, x2, y2 = map(int, results[0].boxes[0].xyxy[0].cpu().numpy())
+    face_crop = img[y1:y2, x1:x2]
+
+    fface_crop = cv.cvtColor(face_crop, cv.COLOR_BGR2RGB)  # Konversi BGR ke RGB
+    
+    face_pil = Image.fromarray(face_crop)
+    face_tensor = transform(face_pil).unsqueeze(0)
+
+    with torch.no_grad():
+        features = efficientnet_model(face_tensor).cpu().numpy()
+
+    pred = pca_xgb.predict(features)[0]
+    pred = 'Wajah Valid' if pred == 0 else 'Wajah Tidak Valid'
+
+    return f"Predicted class: {pred}"
+
+demo = gr.Interface(
+    fn=predict,
+    inputs=gr.Image(sources="webcam", streaming=True),
+    outputs="text",
+    live=True
+)