from __future__ import absolute_import

from __future__ import division

from __future__ import print_function

import argparse

import os.path

import sys

import numpy as np

from six.moves import xrange

import tensorflow as tf

import new_features_input

import models

from tensorflow.python.platform import gfile

from tensorflow.python.framework import graph_util

#from confuse_test import *

import matplotlib.pyplot as plt

from numpy import asfarray

import rebuild_model

#data_dir = 'tmp/speech_dataset_raw/'

data_dir = '/home/zhangs/tensorflow-master/tensorflow/examples/speech_lyc/tmp/newfeature4/'

#data_dir ='/home/zhangs/tensorflow-master/tensorflow/examples/speech_lyc/tmp/H5_diff_feature/'

silence_percentage = 3

validation_percentage = 0

testing_percentage = 10

sample_rate = 16000

batch_size = 100

#wanted_words ='yes,no,up,down,left,right,on,off,stop,go'

wanted_words ='happy'

#wanted_words='five,happy,left,marvin,nine,seven,sheila,six,stop,zero'

model_architecture = 'dscv1'

clip_duration_ms = 998

window_size_ms = 40

window_stride_ms = 20

dct_coefficient_count = 16

#start_checkpoint = '/home/zhangs/tensorflow-master/tensorflow/examples/speech_lyc/tmp/speech_commands_train/hd5_train_2words_8bit_mixup_16fea_bnnv2/fixed_point_bnn_v2/fixed_point_bnn_v2.ckpt-9800'

start_checkpoint = '/home/zhangs/tensorflow-master/tensorflow/examples/speech_lyc/tmp/speech_commands_train/h5_40_20ms16_1word_happy_dscv1_mix1Q6nobias/dscv1/dscv1.ckpt-5800'

pbs_path = ''

# We want to see all the logging messages for this tutorial.

tf.logging.set_verbosity(tf.logging.INFO)

# Start a new TensorFlow session.

sess = tf.InteractiveSession()

model_settings = models.prepare_model_settings(

len(new_features_input.prepare_words_list_my(wanted_words.split(','))),

sample_rate, clip_duration_ms, window_size_ms,

window_stride_ms, dct_coefficient_count)

audio_processor = new_features_input.AudioProcessor(

data_dir, silence_percentage,

wanted_words.split(','), validation_percentage,

testing_percentage)

fingerprint_size = model_settings['fingerprint_size']

label_count = model_settings['label_count']

fingerprint_input = tf.placeholder(

# tf.float32, [None, fingerprint_size], name='fingerprint_input') #

tf.float32, [None, 20,7,64], name='fingerprint_input') #

ground_truth_input = tf.placeholder(

tf.float32, [None, label_count], name='groundtruth_input')

###########################

dscv1model = rebuild_model.paras(start_checkpoint)

logits,lconv1,lconv1bn,ldsc1,ldsc1bn,ldsc1pw,ldsc1pwbn,ldsc2,ldsc2bn,ldsc2pw,ldsc2pwbn,ldsc3,ldsc3bn,ldsc3pw,ldsc3pwbn,\

ldsc4,ldsc4bn,ldsc4pw,ldsc4pwbn,lfc= dscv1model.build_model(

fingerprint_input,

model_settings)

##########################

softmax = tf.nn.softmax(logits, name='labels_softmax')

print(softmax)

with tf.name_scope('cross_entropy'):

cross_entropy_mean = tf.reduce_mean(

tf.nn.softmax_cross_entropy_with_logits(

labels=ground_truth_input, logits=logits))

predicted_indices = tf.argmax(logits, 1)

expected_indices = tf.argmax(ground_truth_input, 1)

correct_prediction = tf.equal(predicted_indices, expected_indices)

confusion_matrix = tf.confusion_matrix(

expected_indices, predicted_indices, num_classes=label_count)

evaluation_step = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

#*********************************************************************

print(" ***************** audio processor ********************")

training_datas = len(audio_processor.data_index['training']) + len(audio_processor.unknown_index['training'])

validation_datas = len(audio_processor.data_index['validation']) + len(audio_processor.unknown_index['validation'])

testing_datas = len(audio_processor.data_index['testing']) + len(audio_processor.unknown_index['testing'])

print("* total samples : " + str(training_datas+validation_datas + testing_datas))

print("* training samples : "+str(len(audio_processor.data_index['training'])) + ' + ' \

+ str(len(audio_processor.unknown_index['training'])) + '(unknowns)' + ' = ' + str(training_datas))

print("* validation samples : "+str(len(audio_processor.data_index['validation']))+ ' + ' \

+ str(len(audio_processor.unknown_index['validation']))+ ' (unknowns)' + ' = ' + str(validation_datas))

print("* testing samples : "+str(len(audio_processor.data_index['testing'])) + ' + ' \

+ str(len(audio_processor.unknown_index['testing'])) + ' (unknowns)' + ' = ' + str(testing_datas))

print(" ********************************************************" + '\n')

#*********************************************************************

print(" *************** Features generator *******************")

test_fingerprints_raw, test_ground_truth= audio_processor.get_data_my(

-1, 0, model_settings, 'testing')

#####################

# test_fingerprints 就是输入数据 维度不记得了，，反正是numpy格式的，你在这里写第一层，然后输出名字还叫这个test_fingerprints，我在model里改输入

dd1=1320

dd4=1

dd2=48

dd3=16

print(test_fingerprints_raw.shape)

input1=test_fingerprints_raw.reshape([1320,48,16])

print(input1)

#dd2,dd3=test_fingerprints.shape

print(dd2,dd3)

wd1,wd2,wd3,wd4=dscv1model.conv1_w.shape

print(wd1,wd2,wd3,wd4)

m=int((dd2-wd1)/2+1)

n=int((dd3-wd2)/2+1)

tmp=np.zeros((dd1,m,n,wd4))

for i in range(dd1):

for j in range(wd4):

for k in range(m):

k1=k*2

for h in range(n):

h1=h*2

temp=np.multiply(input1[i,k1:k1+wd1,h1:h1+wd2],dscv1model.conv1_w[:,:,0,j])

tmp[i,k,h,j]=temp.sum()

test_fingerprints=tmp

#test_fingerprints=test_fingerprints.reshape(dd1,768)

print(test_fingerprints.shape)

##################

print("* fingerprint size ； " + str(model_settings['fingerprint_size']))

print("* test set examples number ； " + str(np.sum(np.sum(test_ground_truth, axis=0))))

print("* test set features size ； " + str(test_fingerprints.shape))

print("* test set labels size ； " + str(test_ground_truth.shape))

if model_settings['fingerprint_size']==test_fingerprints.shape[1] and \

len(new_features_input.prepare_words_list_my(wanted_words.split(','))) == test_ground_truth.shape[1] :

print("-------------> ALL CORRECT <--------------")

else:

print("-------------> DATA WRONG! <--------------")

print(" ********************************************************" + '\n')

print(" *************** import ckpt *******************")

total_accuracy = 0

total_conf_matrix = None

test_accuracy, conf_matrix ,softmax ,correct_prediction,expected_indices,predicted_indices,logits= sess.run(

[evaluation_step, confusion_matrix, softmax,correct_prediction,expected_indices,predicted_indices,logits],

feed_dict={

fingerprint_input: test_fingerprints,

ground_truth_input: test_ground_truth

})

# num_of_inconfident =0

# print("softmax out:" + str(softmax))

# for ii in range(len(list(softmax))):

# if np.max(softmax[ii])<0.8:

# num_of_inconfident += 1

# print("Inconfident sample: " + str(softmax[ii]))

# print("Inconfident number: " + str(num_of_inconfident))

total_accuracy += test_accuracy

if total_conf_matrix is None:

total_conf_matrix = conf_matrix

else:

total_conf_matrix += conf_matrix

tf.logging.info('Confusion Matrix:\n %s' % (total_conf_matrix))

tf.logging.info('Final test accuracy = %.1f%% (N=%d)' % (total_accuracy * 100,testing_datas))

print(" ********************************************************" + '\n')

print(" *************** gate control *******************")

def gate_control(gate,softmax,expected_indices):

return accuracy_gate,softmax_index

#acc , softmax_index = gate_control(0.5,softmax,expected_indices)

print(" ********************************************************" + '\n')

print(" *************** draw ROC *******************")

def FAFR_caculate(start_label,expected_indices,softmax_index):

return FRR,FAR

#FRR,FAR = FAFR_caculate(label_count,expected_indices,softmax_index)

#print("False Alarm Rate = " + str(FAR*100) + '%')

#print("False Reject Rate = " + str(FRR*100) + '%')

def generate_xy(softmax,expected_indices):

return acc_max,acc_max_gate,auc,tipical_fa,[list(x_plot),list(y_plot)]

def plot_and_save(x_plot,y_plot,save_path):

plt.show()

print(" *************** save fa fr data *******************")

save_path = '/home/zhangs/tensorflow-master/tensorflow/examples/speech_lyc/tmp/fa_fr_data/h5_features/' + model_architecture

acc_max,acc_max_gate,auc,tipical_fa,xy = generate_xy(softmax,expected_indices)

plot_and_save(xy[0],xy[1],save_path)

data_array = np.array(xy)

print(type(data_array))

np.savetxt(save_path + "/fafr_data.csv",data_array,delimiter=',')

print(" *************** max accuracy *******************")

print('* AUC : '+ str(auc))

print('* Max accuracy : '+ str(acc_max*100) + ' %' + ' (%.1f%% raw_acc)'% (total_accuracy * 100))

print('* Gate value at max accuracy : '+ str(acc_max_gate))

print('* 0.5%FA point: : '+ str(tipical_fa*100) + '% FR')

print_line = '* AUC : '+ str(auc) + '\n' + \

'* Max accuracy : '+ str(acc_max*100) + ' %' + ' (%.1f%% raw_acc)'% (total_accuracy * 100) + '\n' + \

'* Gate value at max accuracy : '+ str(acc_max_gate) + '\n' + \

'* 0.5%FA point: : '+ str(tipical_fa*100) + '% FR'

with open (save_path + '/details.txt','w') as f:

f.write(print_line)

num_of_unconfident = 0

num_of_wrong = 0

print("softmax out:" + str(softmax))

_, softmax_index_final = gate_control(acc_max_gate, softmax, expected_indices)

confusion_matrix_final = tf.confusion_matrix(

expected_indices, softmax_index_final, num_classes=label_count)

for ii in range(len(list(softmax))):

if np.max(softmax[ii])<acc_max_gate:

num_of_unconfident += 1

print("Unconfident sample: " + str(softmax[ii]))

#if np.max(softmax[ii]) != softmax[ii][0]:

if expected_indices[ii] != 0:

print("Wrong sample: " + str(softmax[ii]))

num_of_wrong +=1

print("Unconfident number: " + str(num_of_unconfident))

print("Wrong number: " + str(num_of_wrong))

print("confusion matrix:" )

print(str(sess.run(confusion_matrix_final)))