Commit f731b60bffd67122b09f6044600147f00ca59c44
1 parent
163652ee
Exists in
refactor
SVM2
Showing
2 changed files
with
227 additions
and
2 deletions
Show diff stats
| @@ -0,0 +1,225 @@ | @@ -0,0 +1,225 @@ | ||
| 1 | +''' | ||
| 2 | +SVM Model. | ||
| 3 | + | ||
| 4 | +@author: chunk | ||
| 5 | +chunkplus@gmail.com | ||
| 6 | +2014 Dec | ||
| 7 | +''' | ||
| 8 | +import os, sys | ||
| 9 | +# from ...mfeat import * | ||
| 10 | +from ...mmodel import * | ||
| 11 | +# from ...mmodel.svm.svmutil import * | ||
| 12 | +from ...common import * | ||
| 13 | + | ||
| 14 | +import numpy as np | ||
| 15 | +import csv | ||
| 16 | +import json | ||
| 17 | +import pickle | ||
| 18 | +# import cv2 | ||
| 19 | +from sklearn import svm | ||
| 20 | + | ||
| 21 | +package_dir = os.path.dirname(os.path.abspath(__file__)) | ||
| 22 | + | ||
| 23 | +dict_Train = {} | ||
| 24 | +dict_databuf = {} | ||
| 25 | +dict_tagbuf = {} | ||
| 26 | +dict_featbuf = {} | ||
| 27 | + | ||
| 28 | + | ||
| 29 | +class ModelSVM(ModelBase): | ||
| 30 | + def __init__(self, toolset='sklearn', sc=None): | ||
| 31 | + ModelBase.__init__(self) | ||
| 32 | + self.toolset = toolset | ||
| 33 | + self.sparker = sc | ||
| 34 | + | ||
| 35 | + | ||
| 36 | + def _train_sklearn(self, X, Y): | ||
| 37 | + clf = svm.SVC(C=4, kernel='linear', shrinking=False, verbose=True) | ||
| 38 | + clf.fit(X, Y) | ||
| 39 | + with open(os.path.join(package_dir, '../..', 'res/svm_sklearn.model'), 'wb') as modelfile: | ||
| 40 | + model = pickle.dump(clf, modelfile) | ||
| 41 | + | ||
| 42 | + self.model = clf | ||
| 43 | + | ||
| 44 | + return clf | ||
| 45 | + | ||
| 46 | + | ||
| 47 | + def _predict_sklearn(self, feat, model=None): | ||
| 48 | + """N.B. sklearn.svm.base.predict : | ||
| 49 | + Perform classification on samples in X. | ||
| 50 | + Parameters | ||
| 51 | + ---------- | ||
| 52 | + X : {array-like, sparse matrix}, shape = [n_samples, n_features] | ||
| 53 | + | ||
| 54 | + Returns | ||
| 55 | + ------- | ||
| 56 | + y_pred : array, shape = [n_samples] | ||
| 57 | + Class labels for samples in X. | ||
| 58 | + """ | ||
| 59 | + if model is None: | ||
| 60 | + if self.model != None: | ||
| 61 | + model = self.model | ||
| 62 | + else: | ||
| 63 | + print 'loading model ...' | ||
| 64 | + with open(os.path.join(package_dir, '../..', 'res/svm_sklearn.model'), 'rb') as modelfile: | ||
| 65 | + model = pickle.load(modelfile) | ||
| 66 | + | ||
| 67 | + return model.predict(feat) | ||
| 68 | + | ||
| 69 | + def __test_sklearn(self, X, Y, model=None): | ||
| 70 | + if model is None: | ||
| 71 | + if self.model != None: | ||
| 72 | + model = self.model | ||
| 73 | + else: | ||
| 74 | + print 'loading model ...' | ||
| 75 | + with open(os.path.join(package_dir, '../..', 'res/svm_sklearn.model'), 'rb') as modelfile: | ||
| 76 | + model = pickle.load(modelfile) | ||
| 77 | + | ||
| 78 | + result_Y = np.array(self._predict_sklearn(X, model)) | ||
| 79 | + | ||
| 80 | + fp = 0 | ||
| 81 | + tp = 0 | ||
| 82 | + sum = np.sum(np.array(Y) == 1) | ||
| 83 | + positive, negative = np.sum(np.array(Y) == 1), np.sum(np.array(Y) == 0) | ||
| 84 | + print positive, negative | ||
| 85 | + for i in range(len(Y)): | ||
| 86 | + if Y[i] == 0 and result_Y[i] == 1: | ||
| 87 | + fp += 1 | ||
| 88 | + elif Y[i] == 1 and result_Y[i] == 1: | ||
| 89 | + tp += 1 | ||
| 90 | + return float(fp) / negative, float(tp) / positive, np.mean(Y == result_Y) | ||
| 91 | + | ||
| 92 | + def _test_sklearn(self, X, Y, model=None): | ||
| 93 | + if model is None: | ||
| 94 | + if self.model != None: | ||
| 95 | + model = self.model | ||
| 96 | + else: | ||
| 97 | + print 'loading model ...' | ||
| 98 | + with open(os.path.join(package_dir, '../..', 'res/svm_sklearn.model'), 'rb') as modelfile: | ||
| 99 | + model = pickle.load(modelfile) | ||
| 100 | + | ||
| 101 | + return model.score(X, Y) | ||
| 102 | + | ||
| 103 | + # def _train_libsvm(self, X, Y): | ||
| 104 | + # X, Y = list(X), list(Y) | ||
| 105 | + # # X, Y = [float(i) for i in X], [float(i) for i in Y] | ||
| 106 | + # prob = svm_problem(Y, X) | ||
| 107 | + # param = svm_parameter('-t 0 -c 4 -b 1 -h 0') | ||
| 108 | + # # param = svm_parameter(kernel_type=LINEAR, C=10) | ||
| 109 | + # m = svm_train(prob, param) | ||
| 110 | + # svm_save_model(os.path.join(package_dir, '../..', 'res/svm_libsvm.model'), m) | ||
| 111 | + # | ||
| 112 | + # self.model = m | ||
| 113 | + # | ||
| 114 | + # return m | ||
| 115 | + # | ||
| 116 | + # def _predict_libsvm(self, feat, model=None): | ||
| 117 | + # if model is None: | ||
| 118 | + # if self.model != None: | ||
| 119 | + # model = self.model | ||
| 120 | + # else: | ||
| 121 | + # print 'loading model ...' | ||
| 122 | + # model = svm_load_model(os.path.join(package_dir, '../..', 'res/svm_libsvm.model')) | ||
| 123 | + # | ||
| 124 | + # feat = [list(feat)] | ||
| 125 | + # # print len(feat),[0] * len(feat) | ||
| 126 | + # label, _, _ = svm_predict([0] * len(feat), feat, model) | ||
| 127 | + # return label | ||
| 128 | + # | ||
| 129 | + # | ||
| 130 | + # def _test_libsvm(self, X, Y, model=None): | ||
| 131 | + # if model is None: | ||
| 132 | + # if self.model != None: | ||
| 133 | + # model = self.model | ||
| 134 | + # else: | ||
| 135 | + # print 'loading model ...' | ||
| 136 | + # model = svm_load_model(os.path.join(package_dir, '../..', 'res/svm_libsvm.model')) | ||
| 137 | + # | ||
| 138 | + # X, Y = list(X), list(Y) | ||
| 139 | + # p_labs, p_acc, p_vals = svm_predict(Y, X, model) | ||
| 140 | + # # ACC, MSE, SCC = evaluations(Y, p_labs) | ||
| 141 | + # | ||
| 142 | + # return p_acc | ||
| 143 | + | ||
| 144 | + # def _train_opencv(self, X, Y): | ||
| 145 | + # svm_params = dict(kernel_type=cv2.SVM_LINEAR, | ||
| 146 | + # svm_type=cv2.SVM_C_SVC, | ||
| 147 | + # C=4) | ||
| 148 | + # | ||
| 149 | + # X, Y = np.array(X, dtype=np.float32), np.array(Y, dtype=np.float32) | ||
| 150 | + # | ||
| 151 | + # svm = cv2.SVM() | ||
| 152 | + # svm.train(X, Y, params=svm_params) | ||
| 153 | + # svm.save(os.path.join(package_dir, '../..', 'res/svm_opencv.model')) | ||
| 154 | + # | ||
| 155 | + # self.model = svm | ||
| 156 | + # | ||
| 157 | + # return svm | ||
| 158 | + # | ||
| 159 | + # | ||
| 160 | + # def _predict_opencv(self, feat, model=None): | ||
| 161 | + # if model is None: | ||
| 162 | + # if self.model != None: | ||
| 163 | + # model = self.model | ||
| 164 | + # else: | ||
| 165 | + # print 'loading model ...' | ||
| 166 | + # with open(os.path.join(package_dir, '../..', 'res/svm_opencv.model'), 'rb') as modelfile: | ||
| 167 | + # model = pickle.load(modelfile) | ||
| 168 | + # feat = np.array(feat, dtype=np.float32) | ||
| 169 | + # | ||
| 170 | + # return model.predict(feat) | ||
| 171 | + # | ||
| 172 | + # | ||
| 173 | + # def _test_opencv(self, X, Y, model=None): | ||
| 174 | + # if model is None: | ||
| 175 | + # if self.model != None: | ||
| 176 | + # model = self.model | ||
| 177 | + # else: | ||
| 178 | + # print 'loading model ...' | ||
| 179 | + # with open(os.path.join(package_dir, '../..', 'res/svm_opencv.model'), 'rb') as modelfile: | ||
| 180 | + # model = pickle.load(modelfile) | ||
| 181 | + # | ||
| 182 | + # X, Y = np.array(X, dtype=np.float32), np.array(Y, dtype=np.float32) | ||
| 183 | + # | ||
| 184 | + # # result_Y = np.array([self._predict_opencv(x, model) for x in X]) | ||
| 185 | + # result_Y = np.array(model.predict_all(X)).ravel() | ||
| 186 | + # # print X[0] | ||
| 187 | + # # print result_Y,Y | ||
| 188 | + # return np.mean(Y == result_Y) | ||
| 189 | + | ||
| 190 | + def train(self, X, Y=None): | ||
| 191 | + | ||
| 192 | + if self.toolset == 'sklearn': | ||
| 193 | + return self._train_sklearn(X, Y) | ||
| 194 | + else: | ||
| 195 | + raise Exception("Unknown toolset!") | ||
| 196 | + | ||
| 197 | + def predict(self, feat, model=None): | ||
| 198 | + | ||
| 199 | + if self.toolset == 'sklearn': | ||
| 200 | + return self._predict_sklearn(feat, model) | ||
| 201 | + else: | ||
| 202 | + raise Exception("Unknown toolset!") | ||
| 203 | + | ||
| 204 | + | ||
| 205 | + def test(self, X, Y=None, model=None): | ||
| 206 | + | ||
| 207 | + if self.toolset == 'sklearn': | ||
| 208 | + return self.__test_sklearn(X, Y, model) | ||
| 209 | + else: | ||
| 210 | + raise Exception("Unknown toolset!") | ||
| 211 | + | ||
| 212 | + | ||
| 213 | + | ||
| 214 | + | ||
| 215 | + | ||
| 216 | + | ||
| 217 | + | ||
| 218 | + | ||
| 219 | + | ||
| 220 | + | ||
| 221 | + | ||
| 222 | + | ||
| 223 | + | ||
| 224 | + | ||
| 225 | + |
mspark/rdd.py
| @@ -6,7 +6,7 @@ from ..mjpeg import * | @@ -6,7 +6,7 @@ from ..mjpeg import * | ||
| 6 | from ..msteg import * | 6 | from ..msteg import * |
| 7 | from ..msteg.steganography import LSB, F3, F4, F5 | 7 | from ..msteg.steganography import LSB, F3, F4, F5 |
| 8 | from ..mfeat import IntraBlockDiff | 8 | from ..mfeat import IntraBlockDiff |
| 9 | -from ..mmodel.svm import SVM | 9 | +from ..mmodel.svm import SVM2 |
| 10 | 10 | ||
| 11 | from numpy import array | 11 | from numpy import array |
| 12 | import json | 12 | import json |
| @@ -19,7 +19,7 @@ from hashlib import md5 | @@ -19,7 +19,7 @@ from hashlib import md5 | ||
| 19 | 19 | ||
| 20 | np.random.seed(sum(map(ord, "whoami"))) | 20 | np.random.seed(sum(map(ord, "whoami"))) |
| 21 | package_dir = os.path.dirname(os.path.abspath(__file__)) | 21 | package_dir = os.path.dirname(os.path.abspath(__file__)) |
| 22 | -classifier = SVM.ModelSVM(toolset='sklearn') | 22 | +classifier = SVM2.ModelSVM(toolset='sklearn') |
| 23 | 23 | ||
| 24 | def rddparse_data_CV(raw_row): | 24 | def rddparse_data_CV(raw_row): |
| 25 | """ | 25 | """ |