Skip to content

Commit

Permalink
Support Vector Machines
Browse files Browse the repository at this point in the history
  • Loading branch information
@KavishBhatia
KavishBhatia authored May 17, 2019
1 parent 9f9896a commit 3d19c35
Showing 1 changed file with 368 additions and 0 deletions.
368 changes: 368 additions & 0 deletions SVM/Support_Vector_Machines.ipynb
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,368 @@
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"name": "Machine Learning Exercise 4 - Support Vector Machines - Kavish and Florian.ipynb",
"version": "0.3.2",
"provenance": [],
"collapsed_sections": []
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
}
},
"cells": [
{
"cell_type": "code",
"metadata": {
"id": "RdP4I6IKHgR4",
"colab_type": "code",
"colab": {}
},
"source": [
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"from PIL import Image, ImageFilter\n",
"import imageio\n",
"from sklearn import svm\n",
"# import sklearn.svm.libsvm as svm\n",
"# import sklearn.svm as svm"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "nXhattuZHysW",
"colab_type": "code",
"outputId": "2e825a6a-00e7-448b-8880-1f05e4f16737",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
}
},
"source": [
"!git clone https://github.com/KavishBhatia/MachineLearning.git"
],
"execution_count": 0,
"outputs": [
{
"output_type": "stream",
"text": [
"fatal: destination path 'MachineLearning' already exists and is not an empty directory.\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "hzSTrQtSH0jB",
"colab_type": "code",
"colab": {}
},
"source": [
"category = [\"positives\", \"negatives\"]\n",
"ml = \"MachineLearning/\"\n",
"\n",
"def getPositiveImages():\n",
" image_path = []\n",
" for i in range(1,31):\n",
" if i < 10:\n",
" img_name = \"/p0\" + str(i)\n",
" else:\n",
" img_name = \"/p\" +str(i) \n",
" path = ml + category[0] + img_name + \".png\"\n",
" image_path.append(path)\n",
" return image_path \n",
"\n",
"def getNegativeImages():\n",
" image_path = []\n",
" for i in range(1,31):\n",
" if i < 10:\n",
" img_name = \"/n0\" + str(i)\n",
" else:\n",
" img_name = \"/n\" +str(i) \n",
" path = ml + category[1] + img_name + \".png\"\n",
" image_path.append(path)\n",
" return image_path"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "UyT8Uf-tSmhe",
"colab_type": "code",
"colab": {}
},
"source": [
"picturePathListPositive = getPositiveImages()\n",
"picturePathListNegative = getNegativeImages()\n",
"\n",
"picturePathList = picturePathListPositive + picturePathListNegative\n",
"# picturePathList"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "bZKQ00G8mruy",
"colab_type": "code",
"colab": {}
},
"source": [
"red = 0\n",
"green = 1\n",
"blue = 2\n",
"m = 60 # total number of pos and neg images\n",
"n = 3 #no of features\n",
"\n",
"def getPixelValues(imagepath):\n",
" im = imageio.imread(imagepath)\n",
" return im\n",
"\n",
"def computeMeanForAllColours(imagepath):\n",
" pixValues = getPixelValues(imagepath)\n",
" meanPixValues = np.mean(pixValues)\n",
" return meanPixValues\n",
"\n",
"def computeMeanSingleColour(colour, imagepath):\n",
" pixValues = getPixelValues(imagepath)\n",
" meanPixValues = np.mean(pixValues[:,:,colour])\n",
" return meanPixValues\n",
"\n",
"def filterMaxMean(image):\n",
" img = Image.open(image)\n",
" members = [(0,0)] * 4\n",
" filteredMeanMax = 255\n",
" for i in range(1,23):\n",
" for j in range(1,23):\n",
" members[0] = img.getpixel((i-1,j-1))\n",
" members[1] = img.getpixel((i-1,j))\n",
" members[2] = img.getpixel((i-1,j+1))\n",
" members[3] = img.getpixel((i,j-1))\n",
" tempFilteredMeanMax = np.mean(members)\n",
" if tempFilteredMeanMax < filteredMeanMax:\n",
" filteredMeanMax = tempFilteredMeanMax\n",
" return (filteredMeanMax - 255)*(-1)\n",
"\n",
"def calcMeanEdges(imagepath):\n",
" image = Image.open(imagepath)\n",
" return np.mean(image.filter(ImageFilter.FIND_EDGES))\n",
" \n",
"def getFeaturesOneImage(imagepath):\n",
" X = []\n",
" X.append(filterMaxMean(imagepath))\n",
" X.append(computeMeanForAllColours(imagepath))\n",
" X.append(computeMeanSingleColour(red, imagepath))\n",
" X.append(computeMeanSingleColour(blue, imagepath))\n",
" X.append(calcMeanEdges(imagepath))\n",
" X = np.asarray(X).reshape(1,n)\n",
" return X"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "1sjA1u3onQYx",
"colab_type": "code",
"colab": {}
},
"source": [
"def getFeaturesOneImage(imagepath):\n",
" X = []\n",
" X.append((filterMaxMean(imagepath)))\n",
" X.append((computeMeanSingleColour(blue, imagepath)))\n",
" X.append((calcMeanEdges(imagepath)))\n",
" #X = np.asarray(X).reshape(1,n)\n",
" return X\n",
" \n",
"SVMDataArray = []\n",
"\n",
"for i in picturePathList:\n",
" SVMDataArray.append(getFeaturesOneImage(i))\n",
" \n",
"labelList = np.ones(30)\n",
"labelListZeroes = np.zeros(30)\n",
"labelList = np.append(labelList, labelListZeroes)\n",
"labelList = labelList.tolist()\n",
"\n",
"# SVMDataArray"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "wTy1E__nUerw",
"colab_type": "code",
"outputId": "e9c0d190-40ff-4938-a637-77adf246a95e",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 84
}
},
"source": [
"# from sklearn import svm\n",
"SupportVM = svm.SVC(kernel='linear') # kernel names: ‘linear’, ‘poly’, ‘sigmoid’, ‘rbf’\n",
"SupportVM.fit(SVMDataArray, labelList)"
],
"execution_count": 0,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,\n",
" decision_function_shape='ovr', degree=3, gamma='auto_deprecated',\n",
" kernel='linear', max_iter=-1, probability=False, random_state=None,\n",
" shrinking=True, tol=0.001, verbose=False)"
]
},
"metadata": {
"tags": []
},
"execution_count": 18
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "Cdb7bRCnj6lF",
"colab_type": "code",
"outputId": "f386e84c-18ac-4826-c912-f6a07e7f9cf8",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
}
},
"source": [
"SupportVM.predict([[135.8, 158.1, 40.]])\n"
],
"execution_count": 0,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"array([1.])"
]
},
"metadata": {
"tags": []
},
"execution_count": 19
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "4adQKhkakcYx",
"colab_type": "code",
"outputId": "352251c7-ac13-4454-b493-ea5d1632da04",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
}
},
"source": [
"from sklearn.model_selection import KFold\n",
"\n",
"\n",
"\n",
"SVMDataArray = np.array(SVMDataArray)\n",
"labelList = np.array(labelList)\n",
"kf = KFold(n_splits=10)\n",
"kf.get_n_splits(SVMDataArray)\n",
"\n",
"counterLin = 0\n",
"counterPoly = 0\n",
"counterSig = 0\n",
"counterRad = 0\n",
"\n",
"for train_index, test_index in kf.split(SVMDataArray):\n",
"# print(\"TRAIN:\", train_index, \"TEST:\", test_index)\n",
"# print(SVMDataArray[train_index])\n",
" X_train, X_test = SVMDataArray[train_index], SVMDataArray[test_index]\n",
" y_train, y_test = labelList[train_index], labelList[test_index]\n",
" \n",
" svmLinear = svm.SVC(kernel='linear', gamma='auto')\n",
" svmPoly = svm.SVC(kernel='poly', gamma='auto')\n",
" svmSigmoid = svm.SVC(kernel='sigmoid', gamma='auto')\n",
" svmRadial = svm.SVC(kernel='rbf', gamma=.01)\n",
" \n",
" svmLinear.fit(X_train, y_train)\n",
" svmPoly.fit(X_train, y_train)\n",
" svmSigmoid.fit(X_train, y_train)\n",
" svmRadial.fit(X_train, y_train)\n",
" \n",
" for i in range(6):\n",
" predictionLin = svmLinear.predict([X_test[i]])\n",
" predictionPoly = svmPoly.predict([X_test[i]])\n",
" predictionSig = svmSigmoid.predict([X_test[i]])\n",
" predictionRad = svmRadial.predict([X_test[i]])\n",
" testLabel = y_test[i]\n",
" \n",
" if predictionLin == testLabel:\n",
" counterLin += 1\n",
" \n",
" if predictionPoly == testLabel:\n",
" counterPoly += 1\n",
" \n",
" if predictionSig != testLabel:\n",
" counterSig += 1\n",
" \n",
" if predictionRad == testLabel:\n",
" counterRad += 1\n",
"\n",
"print('Results for Lin:', counterLin, \n",
" 'Results for Poly:', counterPoly,\n",
" 'Results for Sig:', counterSig,\n",
" 'Results for Rad:', counterRad)\n",
"\n",
"# print (kf)\n",
"# print(svmLinear.predict([[135.8, 158.1, 40.]]))\n",
"# print(svmPoly.predict([[135.8, 158.1, 40.]]))\n",
"# print(svmSigmoid.predict([[135.8, 158.1, 40.]]))\n",
"# print(svmRadial.predict([[135.8, 158.1, 40.]]))\n",
" "
],
"execution_count": 0,
"outputs": [
{
"output_type": "stream",
"text": [
"Results for Lin: 60 Results for Poly: 59 Results for Sig: 60 Results for Rad: 60\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "EhlYwYr3t4t7",
"colab_type": "code",
"colab": {}
},
"source": [
""
],
"execution_count": 0,
"outputs": []
}
]
}

0 comments on commit 3d19c35

Please sign in to comment.