eval.py

import numpy as np
import tensorflow as tf
import os
import sys

prj_dir = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))))
print('prj_dir:', prj_dir)
sys.path.append(prj_dir)

from model import SvfOjb
from params import pm
from dataset.test_data_load import get_dataset
from modules import calculate_eer

# ------------------------------------------------------
# 初始化模型
# ------------------------------------------------------

os.environ["CUDA_VISIBLE_DEVICES"] = '6'

# model_file = "E:\Download/train_model_epoch_1630_gs_40749.0"
model_file = "/home/user/tmp/svf/checkpoint/dev/dev_mode_epoch_2020_gs_250480.0_eer_0.025263"
# model_file = "/home/user/tmp/svf/checkpoint_en/dev/dev_mode_epoch_1259_gs_112051.0_eer_0.015000"

pm.batch_size = 20
pm.utterances_per_speaker = 10

sess = tf.Session()
# 初始化模型
svfObj = SvfOjb()

# 初始化推理图
svfObj.build_inference()
# 保存模型
saver = tf.train.Saver(tf.global_variables(), max_to_keep=pm.max_keep_model)
try:
    saver.restore(sess, model_file)

    print('已加载最近训练模型!')
except Exception as e:
    print("无法加载旧模型,训练过程会覆盖旧模型!")
    pass

# mel_feature_dir = r'E:\dataset\speechRecognition\aishell_test_features'
# mel_feature_dir = '/home/user/tmp/tts/dataset/libriSpeech/aishell_test_features'
mel_feature_dir = '/home/user/tmp/tts/dataset/libriSpeech/aishell_train_features'
# mel_feature_dir = '/home/user/tmp/tts/dataset/libriSpeech/test-clean_features'
# mel_feature_dir = '/home/user/tmp/tts/dataset/libriSpeech/train-other-500_features'
# mel_feature_dir = '/home/user/tmp/tts/dataset/libriSpeech/MAGICDATA_Mandarin_Chinese_Speech_test_features'
meta_file = mel_feature_dir + '/meta.txt'
dataset, num_batch = get_dataset(mel_feature_base_dir=mel_feature_dir, meta_file=meta_file)
iterator = dataset.make_initializable_iterator()
next_element = iterator.get_next()
sess.run(iterator.initializer)

# 冻结模型
tf.get_default_graph().finalize()

# ------------------------------------------------------
# 计算测试集的eer
# ------------------------------------------------------

eer_cl = []
for step in range(num_batch * 5):
    feature = sess.run(next_element)

    [sim_matrix] = sess.run([svfObj.sim_matrix],
                            feed_dict={svfObj.inpt_inference: feature})

    eer_v = calculate_eer(sim_matrix)

    print('eer_v:', eer_v)
    eer_cl.append(eer_v)

print('==========avg_eer:', np.mean(np.array(eer_cl)))
print('测试目录:', mel_feature_dir)