基于深度学习的响指识别

部署

1. 准备工作

你需要有一个麦克风

2. 克隆本项目

git clone git@github.com:fujr/snap_detect

3. 准备python环境

python -m venv venv
source venv/bin/activate 
venv\Scripts\activate (Windows)
pip install -r requirements.txt

4. 启动

1. 训练（非必须）：使用record.py录制环境音和你的响指声音，录制完的文件默认在saved_audio目录下，你需要移动到dataset/train下

   python record.py
   cp -r ./saved_audio/* ./dataset/train/
   python train.py
   
   

   训练时，这些参数是可以改动的

   #train.py
   #dataset参数和训练效果导出的文件名
   TRAIN_DATASET_PATH="dataset/train"
   NUM_MEL_BINS = 15
   EPOCHS = 100
   SIMPLE_RATE = 16000
   SECOND = 1
   BATCH_SIZE =32
   FRAME_LENGTH = 2000
   FRAME_STEP = 1000
   FRAME_NUM = int(16000 / FRAME_STEP - (FRAME_LENGTH / FRAME_STEP - 1))
   LOWER_EDGE_HERTZ = 1500
   UPPER_EDGE_HERTZ = 5500
   DATE = time.strftime("%m%d_%H%M", time.localtime())
   DIR = 'model_%s' % DATE
   TF_MODEL = "%s/model_%s.%s"
   TFLITE_MODEL = "%s/model_%s.tflite"
   ROC_PNG = "%s/roc_%s.png"
   MEL_PNG = "%s/mel.png"
   SNAP_C = "%s/snap.c"
   MODEL_INFO = "%s/model_info_%s.json"
   VOICE_TYPE = np.array(tf.io.gfile.listdir(TRAIN_DATASET_PATH))
   TYPE_NUM = len(VOICE_TYPE)
   
   
   #训练过程
   dataset = Dataset(TRAIN_DATASET_PATH)
   train, val, test = dataset.train_ds, dataset.val_ds, dataset.test_ds
   dataset.show(save=True)
   CNN = Sound_Classification_Model('GRU')
   CNN.compile(
       optimizer=tf.keras.optimizers.Adam(learning_rate=1e-3),
       loss=tf.keras.losses.BinaryCrossentropy(),
       metrics=['AUC','accuracy'],
   )
   CNN.fit(train, val, epochs=EPOCHS)
   CNN.evaluate(test,save=True)
   CNN.save(TF_MODEL , save_format='h5')
   LSTM = Sound_Classification_Model('LSTM')
   LSTM.compile(
       optimizer=tf.keras.optimizers.Adam(learning_rate=1e-3),
       loss=tf.keras.losses.BinaryCrossentropy(),
       metrics=['accuracy'],
   )
   LSTM.fit(train, val, epochs=EPOCHS)
   LSTM.evaluate(test,save=True)
   LSTM.save(TF_MODEL , save_format='keras')
   GRU = Sound_Classification_Model('GRU')
   GRU.compile(
       optimizer=tf.keras.optimizers.Adam(learning_rate=1e-3),
       loss=tf.keras.losses.BinaryCrossentropy(),
       metrics=['AUC'],
   )
   GRU.fit(train, val, epochs=EPOCHS)
   GRU.evaluate(test,save=True)
   GRU.save(TF_MODEL , save_format='export')

2. 推理 使用模型进行推理

   python infer.py 模型训练时间 网络类型 模型格式
   #如:3月17日 01：02分训练的LSTM网络 h5格式模型的启动方法为
   python infer.py 0316_0102 LSTM h5