add save error detention somebody to the local folder for retrain

notebook/predict.ipynb

+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import cv2\n",
+    "from sklearn.externals import joblib\n",
+    "from skimage.transform import resize\n",
+    "from sklearn.preprocessing import LabelEncoder\n",
+    "from keras.models import model_from_json\n",
+    "import numpy as np"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 載入人臉偵測cascade分類器\n",
+    "face_cascade = cv2.CascadeClassifier(\"../model/cv2/haarcascade_frontalface_alt2.xml\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 載入Facenet預測模型\n",
+    "model=model_from_json(open(\"../model/keras/facenet_model.json\",\"r\").read())\n",
+    "model.load_weights(\"../model/keras/facenet_weights.h5\")\n",
+    "model.summary()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 載入SVM分類器\n",
+    "clf=joblib.load('../model/20190618224324/20190618224324.pkl')\n",
+    "# 載入LabelEncoder\n",
+    "le=LabelEncoder()\n",
+    "le.classes_ =np.load('../model/20190618224324/classes.npy')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 影像預處理\n",
+    "def prewhiten(x):\n",
+    "    if x.ndim == 4:\n",
+    "        axis = (1, 2, 3)\n",
+    "        size = x[0].size\n",
+    "    elif x.ndim == 3:\n",
+    "        axis = (0, 1, 2)\n",
+    "        size = x.size\n",
+    "    else:\n",
+    "        print(x.ndim)\n",
+    "        raise ValueError('Dimension should be 3 or 4')\n",
+    "\n",
+    "    mean = np.mean(x, axis=axis, keepdims=True)\n",
+    "    std = np.std(x, axis=axis, keepdims=True)\n",
+    "    std_adj = np.maximum(std, 1.0/np.sqrt(size))\n",
+    "    y = (x - mean) / std_adj\n",
+    "    return y\n",
+    "\n",
+    "def l2_normalize(x, axis=-1, epsilon=1e-10):\n",
+    "    output = x / np.sqrt(np.maximum(np.sum(np.square(x), axis=axis, keepdims=True), epsilon))\n",
+    "    return output"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 人臉偵測處理回傳結果矩陣\n",
+    "image_size=160\n",
+    "def face_cropped(img,faces, margin):    \n",
+    "    aligned_images = []\n",
+    "    for f in faces:\n",
+    "        (x, y, w, h) = f\n",
+    "        cropped = img[y-margin//2:y+h+margin//2,x-margin//2:x+w+margin//2, :]\n",
+    "        aligned = resize(cropped, (image_size, image_size), mode='reflect')\n",
+    "        aligned_images.append(aligned)\n",
+    "            \n",
+    "    return np.array(aligned_images)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 取得人臉Facenet預測之特徵值\n",
+    "def calc_embs(faces, margin=10, batch_size=1):\n",
+    "    aligned_images = prewhiten(faces)\n",
+    "    pd = []\n",
+    "    for start in range(0, len(aligned_images), batch_size):\n",
+    "        pd.append(model.predict_on_batch(aligned_images[start:start+batch_size]))\n",
+    "    embs = l2_normalize(np.concatenate(pd))\n",
+    "\n",
+    "    return embs"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 人臉辨識推斷\n",
+    "def infer(le, clf, img):\n",
+    "    faces = face_cascade.detectMultiScale(img,scaleFactor=1.1,minNeighbors=3)\n",
+    "    if(len(faces)==0):\n",
+    "        return '偵測不到人臉請重新調整'\n",
+    "    \n",
+    "    embs = calc_embs(face_cropped(img,faces,10))\n",
+    "#     pred = le.inverse_transform(clf.predict(embs))\n",
+    "    pred=get_labels(le,clf,embs)\n",
+    "    return [faces,pred]\n",
+    "# Labels 解析\n",
+    "def get_labels(le,clf,embs):\n",
+    "    socres=clf.predict_proba(embs)\n",
+    "    print(socres)\n",
+    "    results=[]\n",
+    "    for s in socres:\n",
+    "        if(s[s.argmax()]>0.5):\n",
+    "            results.append(le.inverse_transform([s.argmax()])[0])\n",
+    "        else:\n",
+    "            results.append('Unknow')\n",
+    "    return results"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "img=cv2.imread('../data/Test/Kevin04.jpg')\n",
+    "# img2=cv2.imread('../data/Test/test6.jpg')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "result=infer(le,clf,img)\n",
+    "# result2=infer(le,clf,img2)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 顯示結果\n",
+    "if(len(result)==2):    \n",
+    "    fa=result[0]\n",
+    "    na=result[1]\n",
+    "    for f,n in zip(fa,na):\n",
+    "        x,y,w,h=f\n",
+    "        margin=10\n",
+    "        cv2.rectangle(img,(x-margin//2,y-margin//2),(x+w+margin//2,y+h+margin//2),(0,0,255))\n",
+    "        cv2.putText(img, n, (x-margin//2,y-margin//2), cv2.FONT_HERSHEY_PLAIN,1, (0, 255, 255), 1, cv2.LINE_AA)\n",
+    "    cv2.imshow('frame',img)\n",
+    "    cv2.waitKey(0)\n",
+    "    cv2.destroyAllWindows()\n",
+    "else:\n",
+    "    print(result)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "faces = face_cascade.detectMultiScale(img,scaleFactor=1.1,minNeighbors=3)\n",
+    "if(len(faces)!=0):\n",
+    "    err,embs = calc_embs(face_cropped(faces,10))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "socre=clf.predict_proba(embs)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "get_labels(le,clf,embs)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "fa=result1[0]\n",
+    "na=result1[1]\n",
+    "print(f,n)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "for f,n in zip(fa,na):\n",
+    "    print(str(f) + n)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.5.6"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}