Let’s make sure we are using tensorflow v2+
Training and inference with colab#
Google colab is a free server for running notebooks with GPU/TPU support - this is a great way to use DAS if you do not have a computer with a GPU.
This notebook demonstrates
how to setup DAS
how to load your own datasets
how to train a network and then use that network to label a new audio recording.
Open and edit this notebook in colab by clicking this badge:
%tensorflow_version 2.x
Install DAS:
!pip install das
Collecting das
Downloading das-0.22.1-py3-none-any.whl (77 kB)
|████████████████████████████████| 77 kB 3.0 MB/s eta 0:00:011
?25hRequirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (from das) (1.4.1)
Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from das) (1.19.5)
Collecting zarr
Downloading zarr-2.10.0-py3-none-any.whl (146 kB)
|████████████████████████████████| 146 kB 7.1 MB/s
?25hRequirement already satisfied: librosa in /usr/local/lib/python3.7/dist-packages (from das) (0.8.1)
Requirement already satisfied: pandas in /usr/local/lib/python3.7/dist-packages (from das) (1.1.5)
Collecting flammkuchen
Downloading flammkuchen-0.9.2-py2.py3-none-any.whl (16 kB)
Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from das) (3.2.2)
Requirement already satisfied: scikit-learn in /usr/local/lib/python3.7/dist-packages (from das) (0.22.2.post1)
Collecting peakutils
Downloading PeakUtils-1.3.3-py3-none-any.whl (7.7 kB)
Collecting defopt
Downloading defopt-6.1.0.tar.gz (35 kB)
Collecting matplotlib_scalebar
Downloading matplotlib_scalebar-0.7.2-py2.py3-none-any.whl (17 kB)
Requirement already satisfied: h5py in /usr/local/lib/python3.7/dist-packages (from das) (3.1.0)
Requirement already satisfied: pyyaml in /usr/local/lib/python3.7/dist-packages (from das) (3.13)
Requirement already satisfied: docutils>=0.10 in /usr/local/lib/python3.7/dist-packages (from defopt->das) (0.17.1)
Collecting sphinxcontrib-napoleon>=0.7.0
Downloading sphinxcontrib_napoleon-0.7-py2.py3-none-any.whl (17 kB)
Requirement already satisfied: typing_extensions>=3.7.4 in /usr/local/lib/python3.7/dist-packages (from defopt->das) (3.7.4.3)
Collecting typing_inspect>=0.3.1
Downloading typing_inspect-0.7.1-py3-none-any.whl (8.4 kB)
Requirement already satisfied: six>=1.5.2 in /usr/local/lib/python3.7/dist-packages (from sphinxcontrib-napoleon>=0.7.0->defopt->das) (1.15.0)
Collecting pockets>=0.3
Downloading pockets-0.9.1-py2.py3-none-any.whl (26 kB)
Collecting mypy-extensions>=0.3.0
Downloading mypy_extensions-0.4.3-py2.py3-none-any.whl (4.5 kB)
Requirement already satisfied: tables in /usr/local/lib/python3.7/dist-packages (from flammkuchen->das) (3.4.4)
Requirement already satisfied: cached-property in /usr/local/lib/python3.7/dist-packages (from h5py->das) (1.5.2)
Requirement already satisfied: joblib>=0.14 in /usr/local/lib/python3.7/dist-packages (from librosa->das) (1.0.1)
Requirement already satisfied: soundfile>=0.10.2 in /usr/local/lib/python3.7/dist-packages (from librosa->das) (0.10.3.post1)
Requirement already satisfied: numba>=0.43.0 in /usr/local/lib/python3.7/dist-packages (from librosa->das) (0.51.2)
Requirement already satisfied: decorator>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa->das) (4.4.2)
Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from librosa->das) (21.0)
Requirement already satisfied: pooch>=1.0 in /usr/local/lib/python3.7/dist-packages (from librosa->das) (1.5.1)
Requirement already satisfied: resampy>=0.2.2 in /usr/local/lib/python3.7/dist-packages (from librosa->das) (0.2.2)
Requirement already satisfied: audioread>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa->das) (2.1.9)
Requirement already satisfied: setuptools in /usr/local/lib/python3.7/dist-packages (from numba>=0.43.0->librosa->das) (57.4.0)
Requirement already satisfied: llvmlite<0.35,>=0.34.0.dev0 in /usr/local/lib/python3.7/dist-packages (from numba>=0.43.0->librosa->das) (0.34.0)
Requirement already satisfied: pyparsing>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=20.0->librosa->das) (2.4.7)
Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa->das) (2.23.0)
Requirement already satisfied: appdirs in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa->das) (1.4.4)
Requirement already satisfied: cffi>=1.0 in /usr/local/lib/python3.7/dist-packages (from soundfile>=0.10.2->librosa->das) (1.14.6)
Requirement already satisfied: pycparser in /usr/local/lib/python3.7/dist-packages (from cffi>=1.0->soundfile>=0.10.2->librosa->das) (2.20)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->das) (2.8.2)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->das) (0.10.0)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->das) (1.3.2)
Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/dist-packages (from pandas->das) (2018.9)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->pooch>=1.0->librosa->das) (2.10)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->pooch>=1.0->librosa->das) (3.0.4)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->pooch>=1.0->librosa->das) (2021.5.30)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->pooch>=1.0->librosa->das) (1.24.3)
Requirement already satisfied: numexpr>=2.5.2 in /usr/local/lib/python3.7/dist-packages (from tables->flammkuchen->das) (2.7.3)
Collecting numcodecs>=0.6.4
Downloading numcodecs-0.9.1-cp37-cp37m-manylinux2010_x86_64.whl (6.2 MB)
|████████████████████████████████| 6.2 MB 19.4 MB/s
?25hCollecting fasteners
Downloading fasteners-0.16.3-py2.py3-none-any.whl (28 kB)
Collecting asciitree
Downloading asciitree-0.3.3.tar.gz (4.0 kB)
Building wheels for collected packages: defopt, asciitree
Building wheel for defopt (setup.py) ... ?25l?25hdone
Created wheel for defopt: filename=defopt-6.1.0-py3-none-any.whl size=14367 sha256=e283684ef3edbd107a13b2d88a61e4e1738bc416d8d1fe07af4499c65761f01a
Stored in directory: /root/.cache/pip/wheels/8e/80/07/63d08d3ae3870730bdc3f4d639af4a141f50aa0da27e183912
Building wheel for asciitree (setup.py) ... ?25l?25hdone
Created wheel for asciitree: filename=asciitree-0.3.3-py3-none-any.whl size=5051 sha256=8b134871e2a3d88db6eeeb1d6938976d2892be448eff2ecc74944e9eca70293f
Stored in directory: /root/.cache/pip/wheels/12/1c/38/0def51e15add93bff3f4bf9c248b94db0839b980b8535e72a0
Successfully built defopt asciitree
Installing collected packages: pockets, mypy-extensions, typing-inspect, sphinxcontrib-napoleon, numcodecs, fasteners, asciitree, zarr, peakutils, matplotlib-scalebar, flammkuchen, defopt, das
Successfully installed asciitree-0.3.3 das-0.22.1 defopt-6.1.0 fasteners-0.16.3 flammkuchen-0.9.2 matplotlib-scalebar-0.7.2 mypy-extensions-0.4.3 numcodecs-0.9.1 peakutils-1.3.3 pockets-0.9.1 sphinxcontrib-napoleon-0.7 typing-inspect-0.7.1 zarr-2.10.0
Import all the things:
import das.train, das.predict, das.utils, das.npy_dir
import matplotlib.pyplot as plt
import flammkuchen
import logging
logging.basicConfig(level=logging.INFO)
Mount your google drive so you can access your own datasets - this will ask for authentication.
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
Train the model#
Adjust the variable path_to_data to point to the dataset on your own google drive.
path_to_data = '/content/drive/MyDrive/Dmoj.wrigleyi.npy'
das.train.train(model_name='tcn',
data_dir=path_to_data,
save_dir='res',
nb_hist=1024,
kernel_size=32,
nb_filters=32,
ignore_boundaries=True,
verbose=2,
nb_conv=4,
learning_rate=0.0005,
use_separable=[True, True, False, False],
nb_epoch=1000)
INFO:root:Loading data from /content/drive/MyDrive/Dmoj.wrigleyi.npy.
INFO:root:Version of the data:
INFO:root: MD5 hash of /content/drive/MyDrive/Dmoj.wrigleyi.npy is
INFO:root: 5381c36663f3b7286b0a5c42c0e3e463
INFO:root:Parameters:
INFO:root:{'data_dir': '/content/drive/MyDrive/Dmoj.wrigleyi.npy', 'y_suffix': '', 'save_dir': 'res', 'save_prefix': '', 'model_name': 'tcn', 'nb_filters': 32, 'kernel_size': 32, 'nb_conv': 4, 'use_separable': [True, True, False, False], 'nb_hist': 1024, 'ignore_boundaries': True, 'batch_norm': True, 'nb_pre_conv': 0, 'pre_nb_dft': 64, 'pre_kernel_size': 3, 'pre_nb_filters': 16, 'pre_nb_conv': 2, 'nb_lstm_units': 0, 'verbose': 2, 'batch_size': 32, 'nb_epoch': 1000, 'learning_rate': 0.0005, 'reduce_lr': False, 'reduce_lr_patience': 5, 'fraction_data': None, 'seed': None, 'batch_level_subsampling': False, 'tensorboard': False, 'neptune_api_token': None, 'neptune_project': None, 'log_messages': False, 'nb_stacks': 2, 'with_y_hist': True, 'x_suffix': '', 'balance': False, 'version_data': True, 'sample_weight_mode': 'temporal', 'data_padding': 128, 'return_sequences': True, 'stride': 768, 'y_offset': 0, 'output_stride': 1, 'class_names': ['noise', 'pulse'], 'class_types': ['segment', 'event'], 'filename_endsample_test': [], 'filename_endsample_train': [], 'filename_endsample_val': [], 'filename_startsample_test': [], 'filename_startsample_train': [], 'filename_startsample_val': [], 'filename_train': [], 'filename_val': [], 'samplerate_x_Hz': 10000.0, 'samplerate_y_Hz': 10000.0, 'filename_test': [], 'data_hash': '5381c36663f3b7286b0a5c42c0e3e463', 'nb_freq': 16, 'nb_channels': 16, 'nb_classes': 2, 'first_sample_train': 0, 'last_sample_train': None, 'first_sample_val': 0, 'last_sample_val': None}
INFO:root:Preparing data
INFO:root:Training data:
INFO:root: AudioSequence with 47 batches each with 32 items.
Total of 1158000 samples with
each x=(16,) and
each y=(2,)
INFO:root:Validation data:
INFO:root: AudioSequence with 15 batches each with 32 items.
Total of 386000 samples with
each x=(16,) and
each y=(2,)
INFO:root:building network
/usr/local/lib/python3.7/dist-packages/keras/optimizer_v2/optimizer_v2.py:356: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
"The `lr` argument is deprecated, use `learning_rate` instead.")
INFO:root:None
INFO:root:Will save to res/20210925_132436.
INFO:root:start training
Model: "TCN"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_1 (InputLayer) [(None, 1024, 16)] 0
__________________________________________________________________________________________________
conv1d (Conv1D) (None, 1024, 32) 544 input_1[0][0]
__________________________________________________________________________________________________
separable_conv1d (SeparableConv (None, 1024, 32) 8224 conv1d[0][0]
__________________________________________________________________________________________________
activation (Activation) (None, 1024, 32) 0 separable_conv1d[0][0]
__________________________________________________________________________________________________
lambda (Lambda) (None, 1024, 32) 0 activation[0][0]
__________________________________________________________________________________________________
spatial_dropout1d (SpatialDropo (None, 1024, 32) 0 lambda[0][0]
__________________________________________________________________________________________________
conv1d_1 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d[0][0]
__________________________________________________________________________________________________
add (Add) (None, 1024, 32) 0 conv1d[0][0]
conv1d_1[0][0]
__________________________________________________________________________________________________
separable_conv1d_1 (SeparableCo (None, 1024, 32) 8224 add[0][0]
__________________________________________________________________________________________________
activation_1 (Activation) (None, 1024, 32) 0 separable_conv1d_1[0][0]
__________________________________________________________________________________________________
lambda_1 (Lambda) (None, 1024, 32) 0 activation_1[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_1 (SpatialDro (None, 1024, 32) 0 lambda_1[0][0]
__________________________________________________________________________________________________
conv1d_2 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_1[0][0]
__________________________________________________________________________________________________
add_1 (Add) (None, 1024, 32) 0 add[0][0]
conv1d_2[0][0]
__________________________________________________________________________________________________
separable_conv1d_2 (SeparableCo (None, 1024, 32) 8224 add_1[0][0]
__________________________________________________________________________________________________
activation_2 (Activation) (None, 1024, 32) 0 separable_conv1d_2[0][0]
__________________________________________________________________________________________________
lambda_2 (Lambda) (None, 1024, 32) 0 activation_2[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_2 (SpatialDro (None, 1024, 32) 0 lambda_2[0][0]
__________________________________________________________________________________________________
conv1d_3 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_2[0][0]
__________________________________________________________________________________________________
add_2 (Add) (None, 1024, 32) 0 add_1[0][0]
conv1d_3[0][0]
__________________________________________________________________________________________________
separable_conv1d_3 (SeparableCo (None, 1024, 32) 8224 add_2[0][0]
__________________________________________________________________________________________________
activation_3 (Activation) (None, 1024, 32) 0 separable_conv1d_3[0][0]
__________________________________________________________________________________________________
lambda_3 (Lambda) (None, 1024, 32) 0 activation_3[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_3 (SpatialDro (None, 1024, 32) 0 lambda_3[0][0]
__________________________________________________________________________________________________
conv1d_4 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_3[0][0]
__________________________________________________________________________________________________
add_3 (Add) (None, 1024, 32) 0 add_2[0][0]
conv1d_4[0][0]
__________________________________________________________________________________________________
separable_conv1d_4 (SeparableCo (None, 1024, 32) 8224 add_3[0][0]
__________________________________________________________________________________________________
activation_4 (Activation) (None, 1024, 32) 0 separable_conv1d_4[0][0]
__________________________________________________________________________________________________
lambda_4 (Lambda) (None, 1024, 32) 0 activation_4[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_4 (SpatialDro (None, 1024, 32) 0 lambda_4[0][0]
__________________________________________________________________________________________________
conv1d_5 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_4[0][0]
__________________________________________________________________________________________________
add_4 (Add) (None, 1024, 32) 0 add_3[0][0]
conv1d_5[0][0]
__________________________________________________________________________________________________
separable_conv1d_5 (SeparableCo (None, 1024, 32) 8224 add_4[0][0]
__________________________________________________________________________________________________
activation_5 (Activation) (None, 1024, 32) 0 separable_conv1d_5[0][0]
__________________________________________________________________________________________________
lambda_5 (Lambda) (None, 1024, 32) 0 activation_5[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_5 (SpatialDro (None, 1024, 32) 0 lambda_5[0][0]
__________________________________________________________________________________________________
conv1d_6 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_5[0][0]
__________________________________________________________________________________________________
add_5 (Add) (None, 1024, 32) 0 add_4[0][0]
conv1d_6[0][0]
__________________________________________________________________________________________________
separable_conv1d_6 (SeparableCo (None, 1024, 32) 8224 add_5[0][0]
__________________________________________________________________________________________________
activation_6 (Activation) (None, 1024, 32) 0 separable_conv1d_6[0][0]
__________________________________________________________________________________________________
lambda_6 (Lambda) (None, 1024, 32) 0 activation_6[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_6 (SpatialDro (None, 1024, 32) 0 lambda_6[0][0]
__________________________________________________________________________________________________
conv1d_7 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_6[0][0]
__________________________________________________________________________________________________
add_6 (Add) (None, 1024, 32) 0 add_5[0][0]
conv1d_7[0][0]
__________________________________________________________________________________________________
separable_conv1d_7 (SeparableCo (None, 1024, 32) 8224 add_6[0][0]
__________________________________________________________________________________________________
activation_7 (Activation) (None, 1024, 32) 0 separable_conv1d_7[0][0]
__________________________________________________________________________________________________
lambda_7 (Lambda) (None, 1024, 32) 0 activation_7[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_7 (SpatialDro (None, 1024, 32) 0 lambda_7[0][0]
__________________________________________________________________________________________________
conv1d_8 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_7[0][0]
__________________________________________________________________________________________________
add_7 (Add) (None, 1024, 32) 0 add_6[0][0]
conv1d_8[0][0]
__________________________________________________________________________________________________
separable_conv1d_8 (SeparableCo (None, 1024, 32) 8224 add_7[0][0]
__________________________________________________________________________________________________
activation_8 (Activation) (None, 1024, 32) 0 separable_conv1d_8[0][0]
__________________________________________________________________________________________________
lambda_8 (Lambda) (None, 1024, 32) 0 activation_8[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_8 (SpatialDro (None, 1024, 32) 0 lambda_8[0][0]
__________________________________________________________________________________________________
conv1d_9 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_8[0][0]
__________________________________________________________________________________________________
add_8 (Add) (None, 1024, 32) 0 add_7[0][0]
conv1d_9[0][0]
__________________________________________________________________________________________________
separable_conv1d_9 (SeparableCo (None, 1024, 32) 8224 add_8[0][0]
__________________________________________________________________________________________________
activation_9 (Activation) (None, 1024, 32) 0 separable_conv1d_9[0][0]
__________________________________________________________________________________________________
lambda_9 (Lambda) (None, 1024, 32) 0 activation_9[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_9 (SpatialDro (None, 1024, 32) 0 lambda_9[0][0]
__________________________________________________________________________________________________
conv1d_10 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_9[0][0]
__________________________________________________________________________________________________
add_9 (Add) (None, 1024, 32) 0 add_8[0][0]
conv1d_10[0][0]
__________________________________________________________________________________________________
conv1d_11 (Conv1D) (None, 1024, 32) 32800 add_9[0][0]
__________________________________________________________________________________________________
activation_10 (Activation) (None, 1024, 32) 0 conv1d_11[0][0]
__________________________________________________________________________________________________
lambda_10 (Lambda) (None, 1024, 32) 0 activation_10[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_10 (SpatialDr (None, 1024, 32) 0 lambda_10[0][0]
__________________________________________________________________________________________________
conv1d_12 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_10[0][0]
__________________________________________________________________________________________________
add_10 (Add) (None, 1024, 32) 0 add_9[0][0]
conv1d_12[0][0]
__________________________________________________________________________________________________
conv1d_13 (Conv1D) (None, 1024, 32) 32800 add_10[0][0]
__________________________________________________________________________________________________
activation_11 (Activation) (None, 1024, 32) 0 conv1d_13[0][0]
__________________________________________________________________________________________________
lambda_11 (Lambda) (None, 1024, 32) 0 activation_11[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_11 (SpatialDr (None, 1024, 32) 0 lambda_11[0][0]
__________________________________________________________________________________________________
conv1d_14 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_11[0][0]
__________________________________________________________________________________________________
add_11 (Add) (None, 1024, 32) 0 add_10[0][0]
conv1d_14[0][0]
__________________________________________________________________________________________________
conv1d_15 (Conv1D) (None, 1024, 32) 32800 add_11[0][0]
__________________________________________________________________________________________________
activation_12 (Activation) (None, 1024, 32) 0 conv1d_15[0][0]
__________________________________________________________________________________________________
lambda_12 (Lambda) (None, 1024, 32) 0 activation_12[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_12 (SpatialDr (None, 1024, 32) 0 lambda_12[0][0]
__________________________________________________________________________________________________
conv1d_16 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_12[0][0]
__________________________________________________________________________________________________
add_12 (Add) (None, 1024, 32) 0 add_11[0][0]
conv1d_16[0][0]
__________________________________________________________________________________________________
conv1d_17 (Conv1D) (None, 1024, 32) 32800 add_12[0][0]
__________________________________________________________________________________________________
activation_13 (Activation) (None, 1024, 32) 0 conv1d_17[0][0]
__________________________________________________________________________________________________
lambda_13 (Lambda) (None, 1024, 32) 0 activation_13[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_13 (SpatialDr (None, 1024, 32) 0 lambda_13[0][0]
__________________________________________________________________________________________________
conv1d_18 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_13[0][0]
__________________________________________________________________________________________________
add_13 (Add) (None, 1024, 32) 0 add_12[0][0]
conv1d_18[0][0]
__________________________________________________________________________________________________
conv1d_19 (Conv1D) (None, 1024, 32) 32800 add_13[0][0]
__________________________________________________________________________________________________
activation_14 (Activation) (None, 1024, 32) 0 conv1d_19[0][0]
__________________________________________________________________________________________________
lambda_14 (Lambda) (None, 1024, 32) 0 activation_14[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_14 (SpatialDr (None, 1024, 32) 0 lambda_14[0][0]
__________________________________________________________________________________________________
conv1d_20 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_14[0][0]
__________________________________________________________________________________________________
add_14 (Add) (None, 1024, 32) 0 add_13[0][0]
conv1d_20[0][0]
__________________________________________________________________________________________________
conv1d_21 (Conv1D) (None, 1024, 32) 32800 add_14[0][0]
__________________________________________________________________________________________________
activation_15 (Activation) (None, 1024, 32) 0 conv1d_21[0][0]
__________________________________________________________________________________________________
lambda_15 (Lambda) (None, 1024, 32) 0 activation_15[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_15 (SpatialDr (None, 1024, 32) 0 lambda_15[0][0]
__________________________________________________________________________________________________
conv1d_22 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_15[0][0]
__________________________________________________________________________________________________
add_15 (Add) (None, 1024, 32) 0 add_14[0][0]
conv1d_22[0][0]
__________________________________________________________________________________________________
conv1d_23 (Conv1D) (None, 1024, 32) 32800 add_15[0][0]
__________________________________________________________________________________________________
activation_16 (Activation) (None, 1024, 32) 0 conv1d_23[0][0]
__________________________________________________________________________________________________
lambda_16 (Lambda) (None, 1024, 32) 0 activation_16[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_16 (SpatialDr (None, 1024, 32) 0 lambda_16[0][0]
__________________________________________________________________________________________________
conv1d_24 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_16[0][0]
__________________________________________________________________________________________________
add_16 (Add) (None, 1024, 32) 0 add_15[0][0]
conv1d_24[0][0]
__________________________________________________________________________________________________
conv1d_25 (Conv1D) (None, 1024, 32) 32800 add_16[0][0]
__________________________________________________________________________________________________
activation_17 (Activation) (None, 1024, 32) 0 conv1d_25[0][0]
__________________________________________________________________________________________________
lambda_17 (Lambda) (None, 1024, 32) 0 activation_17[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_17 (SpatialDr (None, 1024, 32) 0 lambda_17[0][0]
__________________________________________________________________________________________________
conv1d_26 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_17[0][0]
__________________________________________________________________________________________________
add_17 (Add) (None, 1024, 32) 0 add_16[0][0]
conv1d_26[0][0]
__________________________________________________________________________________________________
conv1d_27 (Conv1D) (None, 1024, 32) 32800 add_17[0][0]
__________________________________________________________________________________________________
activation_18 (Activation) (None, 1024, 32) 0 conv1d_27[0][0]
__________________________________________________________________________________________________
lambda_18 (Lambda) (None, 1024, 32) 0 activation_18[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_18 (SpatialDr (None, 1024, 32) 0 lambda_18[0][0]
__________________________________________________________________________________________________
conv1d_28 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_18[0][0]
__________________________________________________________________________________________________
add_18 (Add) (None, 1024, 32) 0 add_17[0][0]
conv1d_28[0][0]
__________________________________________________________________________________________________
conv1d_29 (Conv1D) (None, 1024, 32) 32800 add_18[0][0]
__________________________________________________________________________________________________
activation_19 (Activation) (None, 1024, 32) 0 conv1d_29[0][0]
__________________________________________________________________________________________________
lambda_19 (Lambda) (None, 1024, 32) 0 activation_19[0][0]
__________________________________________________________________________________________________
spatial_dropout1d_19 (SpatialDr (None, 1024, 32) 0 lambda_19[0][0]
__________________________________________________________________________________________________
conv1d_30 (Conv1D) (None, 1024, 32) 1056 spatial_dropout1d_19[0][0]
__________________________________________________________________________________________________
add_20 (Add) (None, 1024, 32) 0 conv1d_1[0][0]
conv1d_2[0][0]
conv1d_3[0][0]
conv1d_4[0][0]
conv1d_5[0][0]
conv1d_6[0][0]
conv1d_7[0][0]
conv1d_8[0][0]
conv1d_9[0][0]
conv1d_10[0][0]
conv1d_12[0][0]
conv1d_14[0][0]
conv1d_16[0][0]
conv1d_18[0][0]
conv1d_20[0][0]
conv1d_22[0][0]
conv1d_24[0][0]
conv1d_26[0][0]
conv1d_28[0][0]
conv1d_30[0][0]
__________________________________________________________________________________________________
activation_20 (Activation) (None, 1024, 32) 0 add_20[0][0]
__________________________________________________________________________________________________
dense (Dense) (None, 1024, 2) 66 activation_20[0][0]
__________________________________________________________________________________________________
activation_21 (Activation) (None, 1024, 2) 0 dense[0][0]
==================================================================================================
Total params: 431,970
Trainable params: 431,970
Non-trainable params: 0
__________________________________________________________________________________________________
Epoch 1/1000
/usr/local/lib/python3.7/dist-packages/keras/engine/training.py:2470: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
warnings.warn('`Model.state_updates` will be removed in a future version. '
Epoch 00001: val_loss improved from inf to 0.04569, saving model to res/20210925_132436_model.h5
/usr/local/lib/python3.7/dist-packages/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.
category=CustomMaskWarning)
47/47 - 68s - loss: 0.0796 - val_loss: 0.0457
Epoch 2/1000
Epoch 00002: val_loss improved from 0.04569 to 0.04298, saving model to res/20210925_132436_model.h5
47/47 - 21s - loss: 0.0333 - val_loss: 0.0430
Epoch 3/1000
Epoch 00003: val_loss improved from 0.04298 to 0.04224, saving model to res/20210925_132436_model.h5
47/47 - 21s - loss: 0.0325 - val_loss: 0.0422
Epoch 4/1000
Epoch 00004: val_loss improved from 0.04224 to 0.04108, saving model to res/20210925_132436_model.h5
47/47 - 21s - loss: 0.0317 - val_loss: 0.0411
Epoch 5/1000
Epoch 00005: val_loss did not improve from 0.04108
47/47 - 21s - loss: 0.0328 - val_loss: 0.0416
Epoch 6/1000
Epoch 00006: val_loss improved from 0.04108 to 0.04089, saving model to res/20210925_132436_model.h5
47/47 - 21s - loss: 0.0301 - val_loss: 0.0409
Epoch 7/1000
Epoch 00007: val_loss did not improve from 0.04089
47/47 - 21s - loss: 0.0303 - val_loss: 0.0410
Epoch 8/1000
Epoch 00008: val_loss did not improve from 0.04089
47/47 - 21s - loss: 0.0289 - val_loss: 0.0410
Epoch 9/1000
Epoch 00009: val_loss did not improve from 0.04089
47/47 - 21s - loss: 0.0302 - val_loss: 0.0413
Epoch 10/1000
Epoch 00010: val_loss improved from 0.04089 to 0.04059, saving model to res/20210925_132436_model.h5
47/47 - 21s - loss: 0.0283 - val_loss: 0.0406
Epoch 11/1000
Epoch 00011: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0269 - val_loss: 0.0409
Epoch 12/1000
Epoch 00012: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0279 - val_loss: 0.0410
Epoch 13/1000
Epoch 00013: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0278 - val_loss: 0.0410
Epoch 14/1000
Epoch 00014: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0267 - val_loss: 0.0417
Epoch 15/1000
Epoch 00015: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0267 - val_loss: 0.0409
Epoch 16/1000
Epoch 00016: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0265 - val_loss: 0.0408
Epoch 17/1000
Epoch 00017: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0266 - val_loss: 0.0415
Epoch 18/1000
Epoch 00018: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0271 - val_loss: 0.0424
Epoch 19/1000
Epoch 00019: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0287 - val_loss: 0.0429
Epoch 20/1000
Epoch 00020: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0268 - val_loss: 0.0412
Epoch 21/1000
Epoch 00021: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0272 - val_loss: 0.0423
Epoch 22/1000
Epoch 00022: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0276 - val_loss: 0.0417
Epoch 23/1000
Epoch 00023: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0268 - val_loss: 0.0417
Epoch 24/1000
Epoch 00024: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0271 - val_loss: 0.0418
Epoch 25/1000
Epoch 00025: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0269 - val_loss: 0.0421
Epoch 26/1000
Epoch 00026: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0258 - val_loss: 0.0420
Epoch 27/1000
Epoch 00027: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0259 - val_loss: 0.0421
Epoch 28/1000
Epoch 00028: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0256 - val_loss: 0.0424
Epoch 29/1000
Epoch 00029: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0281 - val_loss: 0.0438
Epoch 30/1000
Epoch 00030: val_loss did not improve from 0.04059
47/47 - 21s - loss: 0.0269 - val_loss: 0.0424
INFO:root:re-loading last best model
INFO:root:predicting
INFO:root:evaluating
INFO:root:[[354461 1856]
[ 1718 10605]]
INFO:root:{'noise': {'precision': 0.9951765825610157, 'recall': 0.9947911550669769, 'f1-score': 0.9949838314881768, 'support': 356317}, 'pulse': {'precision': 0.8510552925126394, 'recall': 0.8605858962914875, 'f1-score': 0.8557940606843125, 'support': 12323}, 'accuracy': 0.9903049045138889, 'macro avg': {'precision': 0.9231159375368276, 'recall': 0.9276885256792322, 'f1-score': 0.9253889460862447, 'support': 368640}, 'weighted avg': {'precision': 0.9903588561686921, 'recall': 0.9903049045138889, 'f1-score': 0.9903309572867444, 'support': 368640}}
INFO:root:saving to res/20210925_132436_results.h5.
/usr/local/lib/python3.7/dist-packages/tables/path.py:112: NaturalNameWarning: object name is not a valid Python identifier: 'f1-score'; it does not match the pattern ``^[a-zA-Z_][a-zA-Z0-9_]*$``; you will not be able to use natural naming to access this object; using ``getattr()`` will still work, though
NaturalNameWarning)
/usr/local/lib/python3.7/dist-packages/tables/path.py:112: NaturalNameWarning: object name is not a valid Python identifier: 'macro avg'; it does not match the pattern ``^[a-zA-Z_][a-zA-Z0-9_]*$``; you will not be able to use natural naming to access this object; using ``getattr()`` will still work, though
NaturalNameWarning)
/usr/local/lib/python3.7/dist-packages/tables/path.py:112: NaturalNameWarning: object name is not a valid Python identifier: 'weighted avg'; it does not match the pattern ``^[a-zA-Z_][a-zA-Z0-9_]*$``; you will not be able to use natural naming to access this object; using ``getattr()`` will still work, though
NaturalNameWarning)
INFO:root:DONE.
(<keras.engine.functional.Functional at 0x7f269f6ed290>,
{'balance': False,
'batch_level_subsampling': False,
'batch_norm': True,
'batch_size': 32,
'class_names': ['noise', 'pulse'],
'class_types': ['segment', 'event'],
'class_weights': None,
'data_dir': '/content/drive/MyDrive/Dmoj.wrigleyi.npy',
'data_hash': '5381c36663f3b7286b0a5c42c0e3e463',
'data_padding': 128,
'filename_endsample_test': [],
'filename_endsample_train': [],
'filename_endsample_val': [],
'filename_startsample_test': [],
'filename_startsample_train': [],
'filename_startsample_val': [],
'filename_test': [],
'filename_train': [],
'filename_val': [],
'first_sample_train': 0,
'first_sample_val': 0,
'fraction_data': None,
'ignore_boundaries': True,
'kernel_size': 32,
'last_sample_train': None,
'last_sample_val': None,
'learning_rate': 0.0005,
'log_messages': False,
'model_name': 'tcn',
'nb_channels': 16,
'nb_classes': 2,
'nb_conv': 4,
'nb_epoch': 1000,
'nb_filters': 32,
'nb_freq': 16,
'nb_hist': 1024,
'nb_lstm_units': 0,
'nb_pre_conv': 0,
'nb_stacks': 2,
'neptune_project': None,
'output_stride': 1,
'pre_kernel_size': 3,
'pre_nb_conv': 2,
'pre_nb_dft': 64,
'pre_nb_filters': 16,
'reduce_lr': False,
'reduce_lr_patience': 5,
'return_sequences': True,
'sample_weight_mode': 'temporal',
'samplerate_x_Hz': 10000.0,
'samplerate_y_Hz': 10000.0,
'save_dir': 'res',
'save_prefix': '',
'seed': None,
'stride': 768,
'tensorboard': False,
'use_separable': [True, True, False, False],
'verbose': 2,
'version_data': True,
'with_y_hist': True,
'x_suffix': '',
'y_offset': 0,
'y_suffix': ''})
Adjust the name to point to the results:
res_name = '/content/res/20210925_132436'
res = flammkuchen.load(f'{res_name}_results.h5')
Inspect the history of the training and validation loss
plt.plot(res['fit_hist']['loss'], label='train')
plt.plot(res['fit_hist']['val_loss'], label='val')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend()
plt.show()
Plot the test results:
# t0, t1 = 1_020_000, 1_040_000 # dmel tutorial dataset
t0, t1 = 40_000, 60_000
plt.figure(figsize=(40, 8))
plt.subplot(311)
plt.plot(res['x_test'][t0:t1], 'k')
plt.ylabel('Audio')
plt.subplot(312)
plt.plot(res['y_test'][t0:t1, 1:])
plt.ylabel('Prediction targets')
plt.subplot(313)
plt.plot(res['y_pred'][t0:t1, 1:])
plt.ylabel('Predictions')
plt.show()
You can download the model results via the file tab on the left, from /contest/res
Predict on new data#
Load a new recording for prediction
model, params = das.utils.load_model_and_params(res_name) # load the model and runtime parameters
ds = das.npy_dir.load('/content/drive/MyDrive/Dmoj.wrigleyi.npy', memmap_dirs=['train','val']) # load the new data
print(ds)
x = ds['test']['x']
Data:
test:
x: (386000, 16)
y: (386000, 2)
val:
y: (386000, 2)
x: (386000, 16)
train:
y: (1158000, 2)
x: (1158000, 16)
Attributes:
class_names: ['noise', 'pulse']
class_types: ['segment', 'event']
filename_endsample_test: []
filename_endsample_train: []
filename_endsample_val: []
filename_startsample_test: []
filename_startsample_train: []
filename_startsample_val: []
filename_train: []
filename_val: []
samplerate_x_Hz: 10000.0
samplerate_y_Hz: 10000.0
filename_test: []
Run inference - this will calculate the confidence score and extract segment boundaries and event times.
events, segments, class_probabilities, _ = das.predict.predict(x, model=model, params=params, verbose=1)
/usr/local/lib/python3.7/dist-packages/keras/engine/training.py:2470: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
warnings.warn('`Model.state_updates` will be removed in a future version. '
t0, t1 = 42_000, 48_000
plt.figure(figsize=(20, 8))
plt.plot(x[t0:t1, 0], alpha=0.25, c='k', label='Audio')
plt.plot(class_probabilities[t0:t1, 1:], label='Prediction')
plt.xlim(0, t1-t0)
plt.legend(['Audio', 'Prediction'])
plt.show()
