def runQueryAnalysis(audioTransName, hashprintTransName, testName):
print("[runQueryAnalysis] start analysis of:", \
audioTransName, hashprintTransName, testName)
testPath = os.path.join(audioTransName, testName)
dbsPath = os.path.join(audioTransName, hashprintTransName)
resultDict = dict()
for artist in cu.listdir(os.path.join(audioTransName, hashprintTransName)):
print("[runQueryAnalysis] start artist:", artist)
artistResult = dict()
hashprintDB = cu.loadHashprintDB(os.path.join(dbsPath, artist))
for songHashprint in cu.listdir(os.path.join(testPath, artist)):
songHashprintPath = os.path.join(testPath, artist, songHashprint)
artistResult[songHashprint] = hp.runHashprintQuery(songHashprintPath, hashprintDB)
resultDict[artist] = artistResult
analysis = ra.ResultAnalyzer(resultDict, audioTransName, hashprintTransName, testName)
analysis.saveAll()
analysis.saveToCommonCSV()
def runTesting(used_dataset, dataset_code):
verbose = False
multiplier = 5
classifiers = 4
metric_file = "classifiers_metrics"
analysis_file = "results"
print("Loading...")
time.sleep(10)
num_instaces = sc._jsc.sc().getExecutorMemoryStatus().size() - 1
print("Instances online: " + str(num_instaces))
mainTime = time.time()
tc.mainTestClassifier(destination_file=metric_file + dataset_code,
verbose=verbose,
multiplier=multiplier,
used_dataset=used_dataset)
print("Test eseguiti in " + str(time.time() - mainTime) + " secondi")
resa.ResultAnalysis(source_file=metric_file + dataset_code,
destination_file=analysis_file + dataset_code,
classifiers=classifiers)
print("Analisi ultimate. File Results pronto")
def calculater_result(model, testdates, testys):
ty = []
pv =[]
print len(testdates)
for i in range(len(testdates)):
pre = bit2num(model.predict(testdates[i]))
pv.append(pre)
tar = bit2num(testys[i][0])
ty.append(tar)
print pre, tar
print ResultAnalysis.SenAndSpe(ty, pv)
break
logger.info('_count = {},_max_accuray]={} '.format(earlystop._count,earlystop._max_accuray))
#testing
logger.info('testing model')
del bl_cnn # deletes the existing model
bl_cnn = load_model(model_file)
score, acc = bl_cnn.evaluate(X_test, y_test, batch_size=50, verbose=1, sample_weight=None)
logger.info('max_val_acc :{}'.format(max_val_acc))
logger.info('Test score:{}'.format( score))
logger.info('Test accuracy: {}'.format(acc))
total_test_acc.append(acc)
#get predict_y
predictions = bl_cnn.predict(X_test)
logger.info('predictions[0:10] = {}'.format(predictions[0:10]))
predict_y = np.ndarray.flatten(np.argmax(predictions, axis= 1 ))
logger.info('predict_y[0:10]={}'.format(predict_y[0:10]))
#export result
index_word = dict([(kv[1],kv[0])for kv in tdg.word_index.items()])
ResultAnalysis.comp_predict_lable(X = X_test,y=y_test,y_predict = predict_y,
index2word=index_word,label2name={0:'negtive',1:'positive'},
file_to_save='./chinese_zi_epoch_{}.txt'.format(k))
logger.info('total_test_acc={}'.format(total_test_acc))
logger.info('mean acc = {}'.format(sum(total_test_acc) / len(total_test_acc)))
logger.info('max acc = {}'.format( max(total_test_acc)))
print(X_test.shape[0], 'test samples')
# 转换为one_hot类型
#Y_test = np_utils.to_categorical(y_test, nb_classes)
Y_test =ones.one_hot_ten(y_test,nb_classes)
print('one-hot-test:',Y_test)
# 构建模型
model = Sequential()
model=load_model('Cnn200.h5')
result=model.predict(X_test)
print(result)
listOne=result[0:28]
#listThree=result[28:]
oneResult=analysisType.resultType(listOne)
#threeResult=analysisType.resultType(listThree)
print('oneResult=',oneResult)
#print('threeResult=',threeResult)
cutPic.cutLoginPicByPicType(oneResult)
##结果分类
# print('total=',len(result))
#
# nb_type=2 #辨识图片种类
# num=len(result)/nb_type
# count=0