Upload 4 files

app (2).py

@@ -0,0 +1,52 @@
+import gradio as gr
+import cv2
+import numpy as np
+from PIL import Image
+import insightface
+import joblib
+
+# ArcFace 얼굴 분석기 로딩
+face_model = insightface.app.FaceAnalysis(name="buffalo_l", providers=['CPUExecutionProvider'])
+face_model.prepare(ctx_id=0)
+
+# SVM 모델 로드
+clf = joblib.load("is_human_classifier.pkl")
+
+# 얼굴 판별 함수
+def predict(image):
+    try:
+        img = np.array(image.convert("RGB"))
+        img_bgr = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+
+        faces = face_model.get(img_bgr)
+        if not faces:
+            return "❌ 얼굴이 감지되지 않았습니다."
+
+        emb = faces[0].embedding.reshape(1, -1)
+        pred = clf.predict(emb)[0]
+        proba = clf.predict_proba(emb)[0][pred]
+        result = "✅ 사람입니다" if pred == 1 else "❌ 사람이 아닙니다"
+        return f"{result} (정확도: {proba:.2f})"
+    except Exception as e:
+        return f"❌ 오류 발생: {str(e)}"
+
+# Gradio 인터페이스 (웹캠 + 업로드 탭 제공)
+with gr.Blocks() as demo:
+    gr.Markdown("## 👤 실시간 얼굴 판별기\n웹캠 또는 이미지 업로드를 통해 사람이 맞는지 판별합니다.")
+    
+    with gr.Tab("📸 Webcam"):
+        webcam_input = gr.Camera(label="웹캠")
+        webcam_output = gr.Textbox(label="판별 결과")
+        webcam_button = gr.Button("예측")
+        webcam_button.click(predict, inputs=webcam_input, outputs=webcam_output)
+
+
+    with gr.Tab("📁 이미지 업로드"):
+        upload_input = gr.Image(label="이미지 업로드")
+        upload_output = gr.Textbox(label="판별 결과")
+        upload_button = gr.Button("예측")
+        upload_button.click(predict, inputs=upload_input, outputs=upload_output)
+
+if __name__ == "__main__":
+    demo.launch()
+

is_human_classifier.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab1a94275b847a7dbbc270b3746f88ae5860dd7e7b9afa340cb1c3e61c30c884
+size 610859

requirements (3).txt

@@ -0,0 +1,7 @@
+gradio>=4.0.0
+opencv-python-headless
+insightface
+onnxruntime
+scikit-learn
+joblib
+Pillow

train_svm.py

@@ -0,0 +1,45 @@
+import os
+import cv2
+import numpy as np
+import insightface
+from sklearn.svm import SVC
+import joblib
+from tqdm import tqdm
+
+# 1. ArcFace 모델 로딩
+model = insightface.app.FaceAnalysis(name="buffalo_l", providers=['CPUExecutionProvider'])
+model.prepare(ctx_id=0)
+
+X = []
+y = []
+
+# 2. 얼굴 임베딩 추출 함수
+def extract_embeddings(folder_path, label):
+    for fname in tqdm(os.listdir(folder_path), desc=f"{label} - {os.path.basename(folder_path)}"):
+        if not fname.lower().endswith((".jpg", ".jpeg", ".png")):
+            continue
+        path = os.path.join(folder_path, fname)
+        img = cv2.imread(path)
+        if img is None:
+            continue
+        faces = model.get(img)
+        if faces:
+            emb = faces[0].embedding
+            X.append(emb)
+            y.append(label)
+
+# 3. 사람(1) / 비사람(0) 폴더로부터 임베딩 추출
+extract_embeddings("dataset/human", 1)
+extract_embeddings("dataset/nonhuman", 0)
+
+print(f"\n✅ 총 샘플 수: {len(X)}")
+print(f"   - 사람 얼굴: {y.count(1)}")
+print(f"   - 비사람: {y.count(0)}")
+
+# 4. SVM 분류기 학습
+clf = SVC(kernel='linear', probability=True)
+clf.fit(X, y)
+
+# 5. 모델 저장
+joblib.dump(clf, "is_human_classifier.pkl")
+print("\n✅ 분류기 저장 완료: is_human_classifier.pkl")