Chunk / ImageR

Download as
Browse Code »

Commit f731b60bffd67122b09f6044600147f00ca59c44

Authored by Chunk
1 parent 163652ee
Exists in refactor

SVM2

Showing 2 changed files with 227 additions and 2 deletions   Show diff stats
mmodel/svm/SVM2.py 0 → 100644
  Diff comments   View file @f731b60
... ... @@ -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
  Diff comments   View file @f731b60
... ... @@ -6,7 +6,7 @@ from ..mjpeg import *
6 6 from ..msteg import *
7 7 from ..msteg.steganography import LSB, F3, F4, F5
8 8 from ..mfeat import IntraBlockDiff
9   -from ..mmodel.svm import SVM
  9 +from ..mmodel.svm import SVM2
10 10  
11 11 from numpy import array
12 12 import json
... ... @@ -19,7 +19,7 @@ from hashlib import md5
19 19  
20 20 np.random.seed(sum(map(ord, "whoami")))
21 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 24 def rddparse_data_CV(raw_row):
25 25 """
... ...