► Code examples / Audio Data / Audio Classification with the STFTSpectrogram layer

Audio Classification with the STFTSpectrogram layer

Author: Mostafa M. Amin
Date created: 2024/10/04
Last modified: 2024/10/04
Description: Introducing the STFTSpectrogram layer to extract spectrograms for audio classification.

ⓘ This example uses Keras 2

View in Colab • GitHub source

Introduction

Preprocessing audio as spectrograms is an essential step in the vast majority of audio-based applications. Spectrograms represent the frequency content of a signal over time, are widely used for this purpose. In this tutorial, we'll demonstrate how to use the STFTSpectrogram layer in Keras to convert raw audio waveforms into spectrograms within the model. We'll then feed these spectrograms into an LSTM network followed by Dense layers to perform audio classification on the Speech Commands dataset.

We will:

  • Load the ESC-10 dataset.
  • Preprocess the raw audio waveforms and generate spectrograms using STFTSpectrogram.
  • Build two models, one using spectrograms as 1D signals and the other is using as images (2D signals) with a pretrained image model.
  • Train and evaluate the models.

Setup

Importing the necessary libraries

import os

os.environ["KERAS_BACKEND"] = "jax"

import keras
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import scipy.io.wavfile
from keras import layers
from scipy.signal import resample

keras.utils.set_random_seed(41)

Define some variables

BASE_DATA_DIR = "./datasets/esc-50_extracted/ESC-50-master/"
BATCH_SIZE = 16
NUM_CLASSES = 10
EPOCHS = 200
SAMPLE_RATE = 16000

Download and Preprocess the ESC-10 Dataset

We'll use the Dataset for Environmental Sound Classification dataset (ESC-10). This dataset consists of five-second .wav files of environmental sounds.

Download and Extract the dataset

keras.utils.get_file(
    "esc-50.zip",
    "https://github.com/karoldvl/ESC-50/archive/master.zip",
    cache_dir="./",
    cache_subdir="datasets",
    extract=True,
)

'./datasets/esc-50_extracted'

Read the CSV file

pd_data = pd.read_csv(os.path.join(BASE_DATA_DIR, "meta", "esc50.csv"))
# filter ESC-50 to ESC-10 and reassign the targets
pd_data = pd_data[pd_data["esc10"]]
targets = sorted(pd_data["target"].unique().tolist())
assert len(targets) == NUM_CLASSES
old_target_to_new_target = {old: new for new, old in enumerate(targets)}
pd_data["target"] = pd_data["target"].map(lambda t: old_target_to_new_target[t])
pd_data
filename fold target category esc10 src_file take
0 1-100032-A-0.wav 1 0 dog True 100032 A
14 1-110389-A-0.wav 1 0 dog True 110389 A
24 1-116765-A-41.wav 1 9 chainsaw True 116765 A
54 1-17150-A-12.wav 1 4 crackling_fire True 17150 A
55 1-172649-A-40.wav 1 8 helicopter True 172649 A
... ... ... ... ... ... ... ...
1876 5-233160-A-1.wav 5 1 rooster True 233160 A
1888 5-234879-A-1.wav 5 1 rooster True 234879 A
1889 5-234879-B-1.wav 5 1 rooster True 234879 B
1894 5-235671-A-38.wav 5 7 clock_tick True 235671 A
1999 5-9032-A-0.wav 5 0 dog True 9032 A

400 rows × 7 columns

Define functions to read and preprocess the WAV files

def read_wav_file(path, target_sr=SAMPLE_RATE):
    sr, wav = scipy.io.wavfile.read(os.path.join(BASE_DATA_DIR, "audio", path))
    wav = wav.astype(np.float32) / 32768.0  # normalize to [-1, 1]
    num_samples = int(len(wav) * target_sr / sr)  # resample to 16 kHz
    wav = resample(wav, num_samples)
    return wav[:, None]  # Add a channel dimension (of size 1)

Create a function that uses the STFTSpectrogram to compute a spectrogram, then plots it.

def plot_single_spectrogram(sample_wav_data):
    spectrogram = layers.STFTSpectrogram(
        mode="log",
        frame_length=SAMPLE_RATE * 20 // 1000,
        frame_step=SAMPLE_RATE * 5 // 1000,
        fft_length=1024,
        trainable=False,
    )(sample_wav_data[None, ...])[0, ...]

    # Plot the spectrogram
    plt.imshow(spectrogram.T, origin="lower")
    plt.title("Single Channel Spectrogram")
    plt.xlabel("Time")
    plt.ylabel("Frequency")
    plt.show()

Create a function that uses the STFTSpectrogram to compute three spectrograms with multiple bandwidths, then aligns them as an image with different channels, to get a multi-bandwith spectrogram, then plots the spectrogram.

def plot_multi_bandwidth_spectrogram(sample_wav_data):
    # All spectrograms must use the same `fft_length`, `frame_step`, and
    # `padding="same"` in order to produce spectrograms with identical shapes,
    # hence aligning them together. `expand_dims` ensures that the shapes are
    # compatible with image models.

    spectrograms = np.concatenate(
        [
            layers.STFTSpectrogram(
                mode="log",
                frame_length=SAMPLE_RATE * x // 1000,
                frame_step=SAMPLE_RATE * 5 // 1000,
                fft_length=1024,
                padding="same",
                expand_dims=True,
            )(sample_wav_data[None, ...])[0, ...]
            for x in [5, 10, 20]
        ],
        axis=-1,
    ).transpose([1, 0, 2])

    # normalize each color channel for better viewing
    mn = spectrograms.min(axis=(0, 1), keepdims=True)
    mx = spectrograms.max(axis=(0, 1), keepdims=True)
    spectrograms = (spectrograms - mn) / (mx - mn)

    plt.imshow(spectrograms, origin="lower")
    plt.title("Multi-bandwidth Spectrogram")
    plt.xlabel("Time")
    plt.ylabel("Frequency")
    plt.show()

Demonstrate a sample wav file.

sample_wav_data = read_wav_file(pd_data["filename"].tolist()[52])
plt.plot(sample_wav_data[:, 0])
plt.show()

png

Plot a Spectrogram

plot_single_spectrogram(sample_wav_data)

png

Plot a multi-bandwidth spectrogram

plot_multi_bandwidth_spectrogram(sample_wav_data)

png

Define functions to construct a TF Dataset

def read_dataset(df, folds):
    msk = df["fold"].isin(folds)
    filenames = df["filename"][msk]
    targets = df["target"][msk].values
    waves = np.array([read_wav_file(fil) for fil in filenames], dtype=np.float32)
    return waves, targets

Create the datasets

train_x, train_y = read_dataset(pd_data, [1, 2, 3])
valid_x, valid_y = read_dataset(pd_data, [4])
test_x, test_y = read_dataset(pd_data, [5])

Training the Models

In this tutorial we demonstrate the different usecases of the STFTSpectrogram layer.

The first model will use a non-trainable STFTSpectrogram layer, so it is intended purely for preprocessing. Additionally, the model will use 1D signals, hence it make use of Conv1D layers.

The second model will use a trainable STFTSpectrogram layer with the expand_dims option, which expands the shapes to be compatible with image models.

Create the 1D model

  1. Create a non-trainable spectrograms, extracting a 1D time signal.
  2. Apply Conv1D layers with LayerNormalization simialar to the classic VGG design.
  3. Apply global maximum pooling to have fixed set of features.
  4. Add Dense layers to make the final predictions based on the features.
model1d = keras.Sequential(
    [
        layers.InputLayer((None, 1)),
        layers.STFTSpectrogram(
            mode="log",
            frame_length=SAMPLE_RATE * 40 // 1000,
            frame_step=SAMPLE_RATE * 15 // 1000,
            trainable=False,
        ),
        layers.Conv1D(64, 64, activation="relu"),
        layers.Conv1D(128, 16, activation="relu"),
        layers.LayerNormalization(),
        layers.MaxPooling1D(4),
        layers.Conv1D(128, 8, activation="relu"),
        layers.Conv1D(256, 8, activation="relu"),
        layers.Conv1D(512, 4, activation="relu"),
        layers.LayerNormalization(),
        layers.Dropout(0.5),
        layers.GlobalMaxPooling1D(),
        layers.Dense(256, activation="relu"),
        layers.Dense(256, activation="relu"),
        layers.Dropout(0.5),
        layers.Dense(NUM_CLASSES, activation="softmax"),
    ],
    name="model_1d_non_trainble_stft",
)
model1d.compile(
    optimizer=keras.optimizers.Adam(1e-5),
    loss="sparse_categorical_crossentropy",
    metrics=["accuracy"],
)
model1d.summary()

Model: "model_1d_non_trainble_stft"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓
┃ Layer (type)                         ┃ Output Shape                ┃         Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩
│ stft_spectrogram_4 (STFTSpectrogram) │ (None, None, 513)           │         656,640 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ conv1d (Conv1D)                      │ (None, None, 64)            │       2,101,312 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ conv1d_1 (Conv1D)                    │ (None, None, 128)           │         131,200 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ layer_normalization                  │ (None, None, 128)           │             256 │
│ (LayerNormalization)                 │                             │                 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ max_pooling1d (MaxPooling1D)         │ (None, None, 128)           │               0 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ conv1d_2 (Conv1D)                    │ (None, None, 128)           │         131,200 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ conv1d_3 (Conv1D)                    │ (None, None, 256)           │         262,400 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ conv1d_4 (Conv1D)                    │ (None, None, 512)           │         524,800 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ layer_normalization_1                │ (None, None, 512)           │           1,024 │
│ (LayerNormalization)                 │                             │                 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ dropout (Dropout)                    │ (None, None, 512)           │               0 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ global_max_pooling1d                 │ (None, 512)                 │               0 │
│ (GlobalMaxPooling1D)                 │                             │                 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ dense (Dense)                        │ (None, 256)                 │         131,328 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ dense_1 (Dense)                      │ (None, 256)                 │          65,792 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ dropout_1 (Dropout)                  │ (None, 256)                 │               0 │
├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤
│ dense_2 (Dense)                      │ (None, 10)                  │           2,570 │
└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘

 Total params: 4,008,522 (15.29 MB)

 Trainable params: 3,351,882 (12.79 MB)

 Non-trainable params: 656,640 (2.50 MB)

Train the model and restore the best weights.

history_model1d = model1d.fit(
    train_x,
    train_y,
    batch_size=BATCH_SIZE,
    validation_data=(valid_x, valid_y),
    epochs=EPOCHS,
    callbacks=[
        keras.callbacks.EarlyStopping(
            monitor="val_loss",
            patience=EPOCHS,
            restore_best_weights=True,
        )
    ],
)

Epoch 1/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 9s 271ms/step - accuracy: 0.1092 - loss: 3.1307 - val_accuracy: 0.0875 - val_loss: 2.4073
Epoch 2/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 2s 6ms/step - accuracy: 0.1434 - loss: 2.6563 - val_accuracy: 0.1000 - val_loss: 2.4051
Epoch 3/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.1324 - loss: 2.5414 - val_accuracy: 0.1000 - val_loss: 2.4050
Epoch 4/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.1552 - loss: 2.4542 - val_accuracy: 0.1000 - val_loss: 2.3832
Epoch 5/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.1204 - loss: 2.3896 - val_accuracy: 0.1000 - val_loss: 2.3405
Epoch 6/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.1210 - loss: 2.3499 - val_accuracy: 0.1000 - val_loss: 2.3108
Epoch 7/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.1547 - loss: 2.2899 - val_accuracy: 0.1000 - val_loss: 2.2994
Epoch 8/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.1672 - loss: 2.2049 - val_accuracy: 0.1250 - val_loss: 2.2802
Epoch 9/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.2025 - loss: 2.1537 - val_accuracy: 0.1000 - val_loss: 2.2709
Epoch 10/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.1832 - loss: 2.1482 - val_accuracy: 0.1500 - val_loss: 2.2698
Epoch 11/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.2389 - loss: 2.0647 - val_accuracy: 0.1000 - val_loss: 2.2354
Epoch 12/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.2253 - loss: 1.9860 - val_accuracy: 0.2125 - val_loss: 2.1661
Epoch 13/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.2123 - loss: 2.0868 - val_accuracy: 0.1125 - val_loss: 2.1726
Epoch 14/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.2390 - loss: 2.0544 - val_accuracy: 0.2375 - val_loss: 2.1123
Epoch 15/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.2656 - loss: 2.0536 - val_accuracy: 0.2625 - val_loss: 2.1235
Epoch 16/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.3263 - loss: 1.9533 - val_accuracy: 0.1750 - val_loss: 2.1477
Epoch 17/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.3790 - loss: 1.8721 - val_accuracy: 0.1875 - val_loss: 2.0823
Epoch 18/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.3292 - loss: 1.8978 - val_accuracy: 0.3125 - val_loss: 2.0181
Epoch 19/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.3430 - loss: 1.8915 - val_accuracy: 0.3625 - val_loss: 1.9877
Epoch 20/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.3613 - loss: 1.7638 - val_accuracy: 0.3500 - val_loss: 1.9599
Epoch 21/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.4141 - loss: 1.6976 - val_accuracy: 0.4125 - val_loss: 1.9317
Epoch 22/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.4173 - loss: 1.6408 - val_accuracy: 0.3000 - val_loss: 1.9310
Epoch 23/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.3887 - loss: 1.5914 - val_accuracy: 0.4500 - val_loss: 1.8504
Epoch 24/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.3943 - loss: 1.5998 - val_accuracy: 0.2875 - val_loss: 1.8993
Epoch 25/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.5392 - loss: 1.4692 - val_accuracy: 0.4000 - val_loss: 1.8548
Epoch 26/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.4735 - loss: 1.5004 - val_accuracy: 0.4250 - val_loss: 1.8440
Epoch 27/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5132 - loss: 1.4321 - val_accuracy: 0.5000 - val_loss: 1.7961
Epoch 28/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5147 - loss: 1.3093 - val_accuracy: 0.4250 - val_loss: 1.8132
Epoch 29/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.5344 - loss: 1.3614 - val_accuracy: 0.5000 - val_loss: 1.7522
Epoch 30/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5545 - loss: 1.2561 - val_accuracy: 0.5375 - val_loss: 1.7180
Epoch 31/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5697 - loss: 1.2651 - val_accuracy: 0.5500 - val_loss: 1.6538
Epoch 32/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5385 - loss: 1.2571 - val_accuracy: 0.6125 - val_loss: 1.6453
Epoch 33/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5734 - loss: 1.3083 - val_accuracy: 0.5125 - val_loss: 1.6801
Epoch 34/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.5976 - loss: 1.1720 - val_accuracy: 0.4625 - val_loss: 1.6860
Epoch 35/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5268 - loss: 1.3844 - val_accuracy: 0.6375 - val_loss: 1.6253
Epoch 36/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.6021 - loss: 1.1720 - val_accuracy: 0.4625 - val_loss: 1.7012
Epoch 37/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5144 - loss: 1.2672 - val_accuracy: 0.6250 - val_loss: 1.5866
Epoch 38/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.6075 - loss: 1.1400 - val_accuracy: 0.6125 - val_loss: 1.5615
Epoch 39/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.6272 - loss: 1.1138 - val_accuracy: 0.5000 - val_loss: 1.6364
Epoch 40/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.5718 - loss: 1.1956 - val_accuracy: 0.6000 - val_loss: 1.6239
Epoch 41/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.5934 - loss: 1.1302 - val_accuracy: 0.5250 - val_loss: 1.5490
Epoch 42/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.5930 - loss: 1.0970 - val_accuracy: 0.5625 - val_loss: 1.5530
Epoch 43/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.6369 - loss: 0.9976 - val_accuracy: 0.6375 - val_loss: 1.5028
Epoch 44/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.6918 - loss: 0.9205 - val_accuracy: 0.6625 - val_loss: 1.4681
Epoch 45/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.6543 - loss: 0.9118 - val_accuracy: 0.6000 - val_loss: 1.4737
Epoch 46/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.6243 - loss: 1.0268 - val_accuracy: 0.5750 - val_loss: 1.5423
Epoch 47/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.6391 - loss: 1.0181 - val_accuracy: 0.6625 - val_loss: 1.4783
Epoch 48/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.6863 - loss: 0.9874 - val_accuracy: 0.7000 - val_loss: 1.3977
Epoch 49/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7209 - loss: 0.8359 - val_accuracy: 0.6625 - val_loss: 1.3844
Epoch 50/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7659 - loss: 0.8241 - val_accuracy: 0.6500 - val_loss: 1.4206
Epoch 51/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7143 - loss: 0.8972 - val_accuracy: 0.6750 - val_loss: 1.3756
Epoch 52/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7081 - loss: 0.9544 - val_accuracy: 0.6375 - val_loss: 1.3703
Epoch 53/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.6907 - loss: 0.9446 - val_accuracy: 0.6750 - val_loss: 1.3564
Epoch 54/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7460 - loss: 0.7399 - val_accuracy: 0.6000 - val_loss: 1.3840
Epoch 55/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7293 - loss: 0.8620 - val_accuracy: 0.6000 - val_loss: 1.3743
Epoch 56/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7504 - loss: 0.7715 - val_accuracy: 0.6875 - val_loss: 1.3175
Epoch 57/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7643 - loss: 0.7617 - val_accuracy: 0.6625 - val_loss: 1.3407
Epoch 58/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7568 - loss: 0.7798 - val_accuracy: 0.6875 - val_loss: 1.2950
Epoch 59/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7863 - loss: 0.6884 - val_accuracy: 0.6625 - val_loss: 1.3306
Epoch 60/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7550 - loss: 0.7504 - val_accuracy: 0.6500 - val_loss: 1.3260
Epoch 61/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8069 - loss: 0.6624 - val_accuracy: 0.6375 - val_loss: 1.3168
Epoch 62/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.7089 - loss: 0.8183 - val_accuracy: 0.7500 - val_loss: 1.2525
Epoch 63/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 7ms/step - accuracy: 0.7407 - loss: 0.7860 - val_accuracy: 0.7000 - val_loss: 1.2101
Epoch 64/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7526 - loss: 0.7691 - val_accuracy: 0.7250 - val_loss: 1.2327
Epoch 65/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7827 - loss: 0.7485 - val_accuracy: 0.6750 - val_loss: 1.2848
Epoch 66/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7195 - loss: 0.7853 - val_accuracy: 0.7000 - val_loss: 1.2047
Epoch 67/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7539 - loss: 0.7530 - val_accuracy: 0.7125 - val_loss: 1.1954
Epoch 68/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7912 - loss: 0.6220 - val_accuracy: 0.6750 - val_loss: 1.2297
Epoch 69/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7688 - loss: 0.6403 - val_accuracy: 0.6375 - val_loss: 1.2524
Epoch 70/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7699 - loss: 0.7181 - val_accuracy: 0.6625 - val_loss: 1.2147
Epoch 71/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.8300 - loss: 0.5858 - val_accuracy: 0.7000 - val_loss: 1.1705
Epoch 72/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 7ms/step - accuracy: 0.7518 - loss: 0.6276 - val_accuracy: 0.7625 - val_loss: 1.1478
Epoch 73/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8192 - loss: 0.5830 - val_accuracy: 0.6750 - val_loss: 1.1484
Epoch 74/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8044 - loss: 0.6725 - val_accuracy: 0.7500 - val_loss: 1.1518
Epoch 75/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7974 - loss: 0.5536 - val_accuracy: 0.6625 - val_loss: 1.2326
Epoch 76/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7249 - loss: 0.7748 - val_accuracy: 0.7500 - val_loss: 1.1622
Epoch 77/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8083 - loss: 0.5952 - val_accuracy: 0.7125 - val_loss: 1.1240
Epoch 78/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8133 - loss: 0.5249 - val_accuracy: 0.7000 - val_loss: 1.1463
Epoch 79/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8088 - loss: 0.5889 - val_accuracy: 0.7375 - val_loss: 1.0684
Epoch 80/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8715 - loss: 0.4484 - val_accuracy: 0.7500 - val_loss: 1.0295
Epoch 81/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.8099 - loss: 0.5720 - val_accuracy: 0.7125 - val_loss: 1.0846
Epoch 82/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8377 - loss: 0.5405 - val_accuracy: 0.7250 - val_loss: 1.0810
Epoch 83/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.7981 - loss: 0.5354 - val_accuracy: 0.7250 - val_loss: 1.0617
Epoch 84/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7894 - loss: 0.5246 - val_accuracy: 0.7625 - val_loss: 1.0503
Epoch 85/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8695 - loss: 0.4168 - val_accuracy: 0.7125 - val_loss: 1.1376
Epoch 86/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.7566 - loss: 0.6546 - val_accuracy: 0.7250 - val_loss: 1.0920
Epoch 87/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8146 - loss: 0.5367 - val_accuracy: 0.6750 - val_loss: 1.0721
Epoch 88/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.8836 - loss: 0.4781 - val_accuracy: 0.7625 - val_loss: 1.0165
Epoch 89/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.8691 - loss: 0.4114 - val_accuracy: 0.7500 - val_loss: 0.9928
Epoch 90/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.8794 - loss: 0.4078 - val_accuracy: 0.7750 - val_loss: 0.9922
Epoch 91/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8698 - loss: 0.4249 - val_accuracy: 0.7375 - val_loss: 1.0113
Epoch 92/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8553 - loss: 0.4388 - val_accuracy: 0.6875 - val_loss: 1.1355
Epoch 93/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8322 - loss: 0.5300 - val_accuracy: 0.7375 - val_loss: 1.0236
Epoch 94/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9123 - loss: 0.4124 - val_accuracy: 0.7625 - val_loss: 0.9826
Epoch 95/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8403 - loss: 0.4664 - val_accuracy: 0.7750 - val_loss: 0.9689
Epoch 96/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8281 - loss: 0.4742 - val_accuracy: 0.7250 - val_loss: 1.1120
Epoch 97/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8416 - loss: 0.4398 - val_accuracy: 0.7375 - val_loss: 1.0888
Epoch 98/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8671 - loss: 0.4704 - val_accuracy: 0.6625 - val_loss: 1.0802
Epoch 99/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.8976 - loss: 0.3859 - val_accuracy: 0.8000 - val_loss: 0.9549
Epoch 100/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8579 - loss: 0.4120 - val_accuracy: 0.7000 - val_loss: 1.0427
Epoch 101/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8420 - loss: 0.4820 - val_accuracy: 0.7500 - val_loss: 0.9615
Epoch 102/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8501 - loss: 0.4540 - val_accuracy: 0.7625 - val_loss: 0.9078
Epoch 103/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8569 - loss: 0.3727 - val_accuracy: 0.6750 - val_loss: 0.9443
Epoch 104/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9123 - loss: 0.2994 - val_accuracy: 0.6875 - val_loss: 0.9821
Epoch 105/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8797 - loss: 0.3424 - val_accuracy: 0.7750 - val_loss: 0.9252
Epoch 106/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8501 - loss: 0.4048 - val_accuracy: 0.7750 - val_loss: 0.9589
Epoch 107/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8604 - loss: 0.3666 - val_accuracy: 0.7375 - val_loss: 0.9306
Epoch 108/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9082 - loss: 0.3093 - val_accuracy: 0.7250 - val_loss: 0.9925
Epoch 109/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.8382 - loss: 0.4424 - val_accuracy: 0.7875 - val_loss: 0.8926
Epoch 110/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9047 - loss: 0.3130 - val_accuracy: 0.7375 - val_loss: 0.9806
Epoch 111/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8886 - loss: 0.3073 - val_accuracy: 0.7375 - val_loss: 0.9880
Epoch 112/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9027 - loss: 0.3040 - val_accuracy: 0.6875 - val_loss: 1.0214
Epoch 113/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8932 - loss: 0.4064 - val_accuracy: 0.7125 - val_loss: 1.0849
Epoch 114/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8624 - loss: 0.4336 - val_accuracy: 0.8000 - val_loss: 0.9287
Epoch 115/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8925 - loss: 0.4030 - val_accuracy: 0.7625 - val_loss: 0.9044
Epoch 116/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.8922 - loss: 0.3145 - val_accuracy: 0.7750 - val_loss: 0.8441
Epoch 117/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9369 - loss: 0.2919 - val_accuracy: 0.7625 - val_loss: 0.8530
Epoch 118/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9051 - loss: 0.2753 - val_accuracy: 0.7250 - val_loss: 0.9205
Epoch 119/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9144 - loss: 0.2948 - val_accuracy: 0.7000 - val_loss: 0.9843
Epoch 120/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9043 - loss: 0.3258 - val_accuracy: 0.7125 - val_loss: 0.9686
Epoch 121/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9383 - loss: 0.2482 - val_accuracy: 0.7125 - val_loss: 0.9158
Epoch 122/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9314 - loss: 0.3248 - val_accuracy: 0.7000 - val_loss: 1.0416
Epoch 123/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8713 - loss: 0.3495 - val_accuracy: 0.7125 - val_loss: 0.9176
Epoch 124/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8660 - loss: 0.3550 - val_accuracy: 0.7750 - val_loss: 0.9248
Epoch 125/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9375 - loss: 0.2040 - val_accuracy: 0.7875 - val_loss: 0.8526
Epoch 126/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9521 - loss: 0.2011 - val_accuracy: 0.7750 - val_loss: 0.8185
Epoch 127/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9070 - loss: 0.2604 - val_accuracy: 0.7875 - val_loss: 0.8706
Epoch 128/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8554 - loss: 0.3367 - val_accuracy: 0.6750 - val_loss: 1.0503
Epoch 129/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8305 - loss: 0.5195 - val_accuracy: 0.7500 - val_loss: 0.9261
Epoch 130/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8939 - loss: 0.3566 - val_accuracy: 0.7875 - val_loss: 0.8478
Epoch 131/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9220 - loss: 0.2700 - val_accuracy: 0.7625 - val_loss: 0.8353
Epoch 132/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.8607 - loss: 0.3409 - val_accuracy: 0.7750 - val_loss: 0.8898
Epoch 133/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8637 - loss: 0.3109 - val_accuracy: 0.7125 - val_loss: 0.9377
Epoch 134/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.8967 - loss: 0.3634 - val_accuracy: 0.7500 - val_loss: 0.9168
Epoch 135/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9148 - loss: 0.2964 - val_accuracy: 0.7250 - val_loss: 0.8667
Epoch 136/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9322 - loss: 0.2350 - val_accuracy: 0.7625 - val_loss: 0.8509
Epoch 137/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9591 - loss: 0.1990 - val_accuracy: 0.8125 - val_loss: 0.7958
Epoch 138/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9115 - loss: 0.2270 - val_accuracy: 0.7250 - val_loss: 0.8488
Epoch 139/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9749 - loss: 0.1524 - val_accuracy: 0.7750 - val_loss: 0.7888
Epoch 140/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9682 - loss: 0.1539 - val_accuracy: 0.8125 - val_loss: 0.7912
Epoch 141/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9379 - loss: 0.1751 - val_accuracy: 0.8125 - val_loss: 0.8002
Epoch 142/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9681 - loss: 0.1103 - val_accuracy: 0.7750 - val_loss: 0.7951
Epoch 143/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9728 - loss: 0.1513 - val_accuracy: 0.7125 - val_loss: 0.8118
Epoch 144/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9460 - loss: 0.1630 - val_accuracy: 0.8125 - val_loss: 0.7843
Epoch 145/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9627 - loss: 0.1494 - val_accuracy: 0.7625 - val_loss: 0.8179
Epoch 146/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9207 - loss: 0.2203 - val_accuracy: 0.7500 - val_loss: 0.8580
Epoch 147/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9507 - loss: 0.1636 - val_accuracy: 0.7875 - val_loss: 0.7897
Epoch 148/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9562 - loss: 0.1523 - val_accuracy: 0.7625 - val_loss: 0.7950
Epoch 149/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9643 - loss: 0.1464 - val_accuracy: 0.7500 - val_loss: 0.8591
Epoch 150/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9449 - loss: 0.1604 - val_accuracy: 0.7250 - val_loss: 0.9112
Epoch 151/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.9043 - loss: 0.2253 - val_accuracy: 0.7875 - val_loss: 0.7553
Epoch 152/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9459 - loss: 0.1466 - val_accuracy: 0.7250 - val_loss: 0.7929
Epoch 153/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9509 - loss: 0.1329 - val_accuracy: 0.8000 - val_loss: 0.7272
Epoch 154/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9458 - loss: 0.2293 - val_accuracy: 0.7500 - val_loss: 0.7482
Epoch 155/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9596 - loss: 0.1434 - val_accuracy: 0.7750 - val_loss: 0.7726
Epoch 156/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9428 - loss: 0.1471 - val_accuracy: 0.8250 - val_loss: 0.7562
Epoch 157/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9775 - loss: 0.1568 - val_accuracy: 0.7625 - val_loss: 0.7586
Epoch 158/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9256 - loss: 0.1936 - val_accuracy: 0.7750 - val_loss: 0.8041
Epoch 159/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9507 - loss: 0.1620 - val_accuracy: 0.7000 - val_loss: 0.9265
Epoch 160/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9545 - loss: 0.2093 - val_accuracy: 0.7875 - val_loss: 0.7786
Epoch 161/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9428 - loss: 0.1747 - val_accuracy: 0.7250 - val_loss: 0.8367
Epoch 162/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9377 - loss: 0.2172 - val_accuracy: 0.7625 - val_loss: 0.7964
Epoch 163/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9509 - loss: 0.1753 - val_accuracy: 0.7500 - val_loss: 0.7437
Epoch 164/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9694 - loss: 0.1197 - val_accuracy: 0.7750 - val_loss: 0.7330
Epoch 165/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9594 - loss: 0.1065 - val_accuracy: 0.7375 - val_loss: 0.8036
Epoch 166/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9752 - loss: 0.1265 - val_accuracy: 0.7000 - val_loss: 0.8316
Epoch 167/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9121 - loss: 0.1863 - val_accuracy: 0.7500 - val_loss: 0.7953
Epoch 168/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9320 - loss: 0.1759 - val_accuracy: 0.8000 - val_loss: 0.8142
Epoch 169/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9613 - loss: 0.1785 - val_accuracy: 0.7625 - val_loss: 0.7585
Epoch 170/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9666 - loss: 0.1096 - val_accuracy: 0.7875 - val_loss: 0.7595
Epoch 171/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9518 - loss: 0.1422 - val_accuracy: 0.7875 - val_loss: 0.7417
Epoch 172/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9689 - loss: 0.1236 - val_accuracy: 0.7625 - val_loss: 0.7539
Epoch 173/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.9959 - loss: 0.0662 - val_accuracy: 0.7875 - val_loss: 0.6840
Epoch 174/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9835 - loss: 0.0803 - val_accuracy: 0.7500 - val_loss: 0.7929
Epoch 175/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9319 - loss: 0.1924 - val_accuracy: 0.7500 - val_loss: 0.8044
Epoch 176/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9290 - loss: 0.2342 - val_accuracy: 0.8000 - val_loss: 0.7280
Epoch 177/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9446 - loss: 0.1692 - val_accuracy: 0.7500 - val_loss: 0.7537
Epoch 178/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9868 - loss: 0.0925 - val_accuracy: 0.8000 - val_loss: 0.7145
Epoch 179/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9788 - loss: 0.1382 - val_accuracy: 0.7625 - val_loss: 0.7860
Epoch 180/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9771 - loss: 0.0829 - val_accuracy: 0.8125 - val_loss: 0.6933
Epoch 181/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9602 - loss: 0.1095 - val_accuracy: 0.7750 - val_loss: 0.7213
Epoch 182/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9723 - loss: 0.1172 - val_accuracy: 0.7500 - val_loss: 0.7286
Epoch 183/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9532 - loss: 0.1564 - val_accuracy: 0.7875 - val_loss: 0.7060
Epoch 184/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 6ms/step - accuracy: 0.9789 - loss: 0.0840 - val_accuracy: 0.8125 - val_loss: 0.6554
Epoch 185/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9857 - loss: 0.0764 - val_accuracy: 0.7875 - val_loss: 0.7785
Epoch 186/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9849 - loss: 0.0791 - val_accuracy: 0.7625 - val_loss: 0.7358
Epoch 187/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9702 - loss: 0.0919 - val_accuracy: 0.7500 - val_loss: 0.7888
Epoch 188/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9931 - loss: 0.0779 - val_accuracy: 0.7625 - val_loss: 0.7874
Epoch 189/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9604 - loss: 0.1247 - val_accuracy: 0.7875 - val_loss: 0.7642
Epoch 190/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9402 - loss: 0.1906 - val_accuracy: 0.7875 - val_loss: 0.8763
Epoch 191/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9845 - loss: 0.1111 - val_accuracy: 0.7875 - val_loss: 0.6824
Epoch 192/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9899 - loss: 0.0591 - val_accuracy: 0.8000 - val_loss: 0.6591
Epoch 193/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9716 - loss: 0.1055 - val_accuracy: 0.7625 - val_loss: 0.7776
Epoch 194/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9750 - loss: 0.0953 - val_accuracy: 0.7250 - val_loss: 0.7947
Epoch 195/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 5ms/step - accuracy: 0.9765 - loss: 0.0889 - val_accuracy: 0.7375 - val_loss: 0.7190
Epoch 196/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9741 - loss: 0.0896 - val_accuracy: 0.8000 - val_loss: 0.7058
Epoch 197/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9586 - loss: 0.0916 - val_accuracy: 0.7625 - val_loss: 0.7676
Epoch 198/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9955 - loss: 0.0655 - val_accuracy: 0.7625 - val_loss: 0.7047
Epoch 199/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9861 - loss: 0.0663 - val_accuracy: 0.7750 - val_loss: 0.7760
Epoch 200/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 4ms/step - accuracy: 0.9982 - loss: 0.0558 - val_accuracy: 0.7750 - val_loss: 0.6585

Create the 2D model

  1. Create three spectrograms with multiple band-widths from the raw input.
  2. Concatenate the three spectrograms to have three channels.
  3. Load MobileNet and set the weights from the weights trained on ImageNet.
  4. Apply global maximum pooling to have fixed set of features.
  5. Add Dense layers to make the final predictions based on the features.
input = layers.Input((None, 1))
spectrograms = [
    layers.STFTSpectrogram(
        mode="log",
        frame_length=SAMPLE_RATE * frame_size // 1000,
        frame_step=SAMPLE_RATE * 15 // 1000,
        fft_length=2048,
        padding="same",
        expand_dims=True,
        # trainable=True,  # trainable by default
    )(input)
    for frame_size in [30, 40, 50]  # frame size in milliseconds
]

multi_spectrograms = layers.Concatenate(axis=-1)(spectrograms)

img_model = keras.applications.MobileNet(include_top=False, pooling="max")
output = img_model(multi_spectrograms)

output = layers.Dropout(0.5)(output)
output = layers.Dense(256, activation="relu")(output)
output = layers.Dense(256, activation="relu")(output)
output = layers.Dense(NUM_CLASSES, activation="softmax")(output)
model2d = keras.Model(input, output, name="model_2d_trainble_stft")

model2d.compile(
    optimizer=keras.optimizers.Adam(1e-4),
    loss="sparse_categorical_crossentropy",
    metrics=["accuracy"],
)
model2d.summary()

<ipython-input-16-bf7092b3c6d2>:17: UserWarning: `input_shape` is undefined or non-square, or `rows` is not in [128, 160, 192, 224]. Weights for input shape (224, 224) will be loaded as the default.
  img_model = keras.applications.MobileNet(include_top=False, pooling="max")

Model: "model_2d_trainble_stft"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Layer (type)              ┃ Output Shape           ┃        Param # ┃ Connected to           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━┩
│ input_layer_1             │ (None, None, 1)        │              0 │ -                      │
│ (InputLayer)              │                        │                │                        │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ stft_spectrogram_5        │ (None, None, 1025, 1)  │        984,000 │ input_layer_1[0][0]    │
│ (STFTSpectrogram)         │                        │                │                        │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ stft_spectrogram_6        │ (None, None, 1025, 1)  │      1,312,000 │ input_layer_1[0][0]    │
│ (STFTSpectrogram)         │                        │                │                        │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ stft_spectrogram_7        │ (None, None, 1025, 1)  │      1,640,000 │ input_layer_1[0][0]    │
│ (STFTSpectrogram)         │                        │                │                        │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ concatenate (Concatenate) │ (None, None, 1025, 3)  │              0 │ stft_spectrogram_5[0]… │
│                           │                        │                │ stft_spectrogram_6[0]… │
│                           │                        │                │ stft_spectrogram_7[0]… │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ mobilenet_1.00_224        │ (None, 1024)           │      3,228,864 │ concatenate[0][0]      │
│ (Functional)              │                        │                │                        │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ dropout_2 (Dropout)       │ (None, 1024)           │              0 │ mobilenet_1.00_224[0]… │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ dense_3 (Dense)           │ (None, 256)            │        262,400 │ dropout_2[0][0]        │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ dense_4 (Dense)           │ (None, 256)            │         65,792 │ dense_3[0][0]          │
├───────────────────────────┼────────────────────────┼────────────────┼────────────────────────┤
│ dense_5 (Dense)           │ (None, 10)             │          2,570 │ dense_4[0][0]          │
└───────────────────────────┴────────────────────────┴────────────────┴────────────────────────┘

 Total params: 7,495,626 (28.59 MB)

 Trainable params: 7,473,738 (28.51 MB)

 Non-trainable params: 21,888 (85.50 KB)

Train the model and restore the best weights.

history_model2d = model2d.fit(
    train_x,
    train_y,
    batch_size=BATCH_SIZE,
    validation_data=(valid_x, valid_y),
    epochs=EPOCHS,
    callbacks=[
        keras.callbacks.EarlyStopping(
            monitor="val_loss",
            patience=EPOCHS,
            restore_best_weights=True,
        )
    ],
)

Epoch 1/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 50s 776ms/step - accuracy: 0.0855 - loss: 7.6484 - val_accuracy: 0.0625 - val_loss: 3.7484
Epoch 2/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 8s 55ms/step - accuracy: 0.1293 - loss: 5.8848 - val_accuracy: 0.0750 - val_loss: 4.0622
Epoch 3/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.1302 - loss: 4.6363 - val_accuracy: 0.0875 - val_loss: 3.6488
Epoch 4/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.1656 - loss: 4.6861 - val_accuracy: 0.1250 - val_loss: 3.5224
Epoch 5/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.2025 - loss: 4.3601 - val_accuracy: 0.0875 - val_loss: 4.0424
Epoch 6/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.2072 - loss: 3.8723 - val_accuracy: 0.1125 - val_loss: 3.1530
Epoch 7/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.2562 - loss: 3.2596 - val_accuracy: 0.1125 - val_loss: 2.9712
Epoch 8/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.2328 - loss: 3.1374 - val_accuracy: 0.1375 - val_loss: 3.0128
Epoch 9/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.3296 - loss: 2.6887 - val_accuracy: 0.1750 - val_loss: 2.6742
Epoch 10/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.3123 - loss: 2.4022 - val_accuracy: 0.1750 - val_loss: 2.7165
Epoch 11/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.3781 - loss: 2.3441 - val_accuracy: 0.1875 - val_loss: 2.1900
Epoch 12/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.4524 - loss: 2.0044 - val_accuracy: 0.3250 - val_loss: 1.8786
Epoch 13/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.3609 - loss: 2.0790 - val_accuracy: 0.3750 - val_loss: 1.7390
Epoch 14/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.5158 - loss: 1.6717 - val_accuracy: 0.3750 - val_loss: 1.5660
Epoch 15/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.5080 - loss: 1.6551 - val_accuracy: 0.4125 - val_loss: 1.6085
Epoch 16/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.5921 - loss: 1.4493 - val_accuracy: 0.5250 - val_loss: 1.2603
Epoch 17/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.5404 - loss: 1.4931 - val_accuracy: 0.6000 - val_loss: 1.0863
Epoch 18/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.6492 - loss: 1.0411 - val_accuracy: 0.6000 - val_loss: 1.0920
Epoch 19/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.5987 - loss: 1.3023 - val_accuracy: 0.5625 - val_loss: 1.0882
Epoch 20/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.5950 - loss: 1.2483 - val_accuracy: 0.5500 - val_loss: 1.0755
Epoch 21/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.5789 - loss: 1.1988 - val_accuracy: 0.5875 - val_loss: 0.9171
Epoch 22/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.6694 - loss: 1.0415 - val_accuracy: 0.6875 - val_loss: 0.8319
Epoch 23/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 53ms/step - accuracy: 0.7705 - loss: 0.8017 - val_accuracy: 0.6750 - val_loss: 0.8824
Epoch 24/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.6693 - loss: 1.0069 - val_accuracy: 0.7500 - val_loss: 0.6454
Epoch 25/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.6997 - loss: 0.8689 - val_accuracy: 0.7250 - val_loss: 0.7640
Epoch 26/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.6816 - loss: 0.8254 - val_accuracy: 0.7500 - val_loss: 0.6418
Epoch 27/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.6524 - loss: 1.1302 - val_accuracy: 0.7375 - val_loss: 0.7160
Epoch 28/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.7624 - loss: 0.7522 - val_accuracy: 0.7875 - val_loss: 0.6805
Epoch 29/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.6926 - loss: 0.8897 - val_accuracy: 0.7500 - val_loss: 0.6289
Epoch 30/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.7190 - loss: 0.7467 - val_accuracy: 0.7375 - val_loss: 0.5838
Epoch 31/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.7171 - loss: 0.7727 - val_accuracy: 0.8250 - val_loss: 0.6101
Epoch 32/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.8120 - loss: 0.5287 - val_accuracy: 0.8625 - val_loss: 0.4229
Epoch 33/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.7921 - loss: 0.5581 - val_accuracy: 0.8250 - val_loss: 0.4174
Epoch 34/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.8056 - loss: 0.5415 - val_accuracy: 0.8500 - val_loss: 0.4672
Epoch 35/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 50ms/step - accuracy: 0.7601 - loss: 0.5661 - val_accuracy: 0.8250 - val_loss: 0.4791
Epoch 36/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.7866 - loss: 0.5135 - val_accuracy: 0.8750 - val_loss: 0.4217
Epoch 37/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.8660 - loss: 0.3952 - val_accuracy: 0.8250 - val_loss: 0.4561
Epoch 38/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.8446 - loss: 0.3751 - val_accuracy: 0.9000 - val_loss: 0.3954
Epoch 39/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.8546 - loss: 0.3984 - val_accuracy: 0.8375 - val_loss: 0.4534
Epoch 40/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.8655 - loss: 0.3541 - val_accuracy: 0.8875 - val_loss: 0.3718
Epoch 41/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.8592 - loss: 0.4164 - val_accuracy: 0.8750 - val_loss: 0.4537
Epoch 42/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9093 - loss: 0.2404 - val_accuracy: 0.8625 - val_loss: 0.4169
Epoch 43/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9329 - loss: 0.1855 - val_accuracy: 0.8750 - val_loss: 0.3354
Epoch 44/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.8353 - loss: 0.4455 - val_accuracy: 0.8750 - val_loss: 0.3619
Epoch 45/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9135 - loss: 0.2196 - val_accuracy: 0.8750 - val_loss: 0.3313
Epoch 46/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9129 - loss: 0.2131 - val_accuracy: 0.8875 - val_loss: 0.3199
Epoch 47/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9467 - loss: 0.1264 - val_accuracy: 0.8875 - val_loss: 0.3162
Epoch 48/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9281 - loss: 0.2276 - val_accuracy: 0.8875 - val_loss: 0.3158
Epoch 49/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9211 - loss: 0.2044 - val_accuracy: 0.8375 - val_loss: 0.3702
Epoch 50/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9247 - loss: 0.1954 - val_accuracy: 0.8750 - val_loss: 0.2875
Epoch 51/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.9534 - loss: 0.1122 - val_accuracy: 0.9000 - val_loss: 0.2637
Epoch 52/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.9596 - loss: 0.1261 - val_accuracy: 0.9125 - val_loss: 0.2370
Epoch 53/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9388 - loss: 0.1679 - val_accuracy: 0.9125 - val_loss: 0.2506
Epoch 54/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9635 - loss: 0.1075 - val_accuracy: 0.9125 - val_loss: 0.2656
Epoch 55/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9511 - loss: 0.1666 - val_accuracy: 0.9000 - val_loss: 0.2998
Epoch 56/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9688 - loss: 0.0860 - val_accuracy: 0.9000 - val_loss: 0.2730
Epoch 57/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9786 - loss: 0.0796 - val_accuracy: 0.8875 - val_loss: 0.2837
Epoch 58/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9421 - loss: 0.1239 - val_accuracy: 0.8750 - val_loss: 0.2829
Epoch 59/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9392 - loss: 0.2626 - val_accuracy: 0.8750 - val_loss: 0.3105
Epoch 60/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9395 - loss: 0.1321 - val_accuracy: 0.9000 - val_loss: 0.2529
Epoch 61/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9679 - loss: 0.0968 - val_accuracy: 0.8750 - val_loss: 0.2506
Epoch 62/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9437 - loss: 0.1074 - val_accuracy: 0.9000 - val_loss: 0.2950
Epoch 63/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9615 - loss: 0.0958 - val_accuracy: 0.8750 - val_loss: 0.3064
Epoch 64/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9755 - loss: 0.0601 - val_accuracy: 0.9000 - val_loss: 0.2795
Epoch 65/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9723 - loss: 0.0673 - val_accuracy: 0.9125 - val_loss: 0.2123
Epoch 66/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.9464 - loss: 0.1619 - val_accuracy: 0.9375 - val_loss: 0.1930
Epoch 67/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9863 - loss: 0.0445 - val_accuracy: 0.9250 - val_loss: 0.1866
Epoch 68/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9823 - loss: 0.0678 - val_accuracy: 0.9125 - val_loss: 0.2109
Epoch 69/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9855 - loss: 0.0579 - val_accuracy: 0.9375 - val_loss: 0.2088
Epoch 70/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.9800 - loss: 0.0549 - val_accuracy: 0.9625 - val_loss: 0.1693
Epoch 71/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9861 - loss: 0.0469 - val_accuracy: 0.9500 - val_loss: 0.1738
Epoch 72/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9876 - loss: 0.0685 - val_accuracy: 0.9375 - val_loss: 0.2090
Epoch 73/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9605 - loss: 0.0835 - val_accuracy: 0.8875 - val_loss: 0.2828
Epoch 74/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9783 - loss: 0.0475 - val_accuracy: 0.8875 - val_loss: 0.2500
Epoch 75/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9871 - loss: 0.0470 - val_accuracy: 0.9000 - val_loss: 0.2094
Epoch 76/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9881 - loss: 0.0405 - val_accuracy: 0.9500 - val_loss: 0.1971
Epoch 77/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 45ms/step - accuracy: 0.9736 - loss: 0.0418 - val_accuracy: 0.9375 - val_loss: 0.2014
Epoch 78/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9582 - loss: 0.1145 - val_accuracy: 0.9125 - val_loss: 0.2082
Epoch 79/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9831 - loss: 0.0586 - val_accuracy: 0.9125 - val_loss: 0.2109
Epoch 80/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9574 - loss: 0.0950 - val_accuracy: 0.9000 - val_loss: 0.3043
Epoch 81/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9964 - loss: 0.0253 - val_accuracy: 0.9250 - val_loss: 0.2476
Epoch 82/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9838 - loss: 0.0427 - val_accuracy: 0.9125 - val_loss: 0.2480
Epoch 83/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0094 - val_accuracy: 0.9250 - val_loss: 0.2614
Epoch 84/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9929 - loss: 0.0256 - val_accuracy: 0.9250 - val_loss: 0.2504
Epoch 85/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9953 - loss: 0.0215 - val_accuracy: 0.9250 - val_loss: 0.2334
Epoch 86/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9939 - loss: 0.0200 - val_accuracy: 0.9500 - val_loss: 0.2138
Epoch 87/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0133 - val_accuracy: 0.9500 - val_loss: 0.2167
Epoch 88/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9907 - loss: 0.0303 - val_accuracy: 0.9125 - val_loss: 0.2326
Epoch 89/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9883 - loss: 0.0406 - val_accuracy: 0.9500 - val_loss: 0.2000
Epoch 90/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9932 - loss: 0.0292 - val_accuracy: 0.9375 - val_loss: 0.1961
Epoch 91/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9756 - loss: 0.1435 - val_accuracy: 0.9375 - val_loss: 0.2093
Epoch 92/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9762 - loss: 0.0868 - val_accuracy: 0.9375 - val_loss: 0.2081
Epoch 93/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9925 - loss: 0.0391 - val_accuracy: 0.9375 - val_loss: 0.1890
Epoch 94/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9961 - loss: 0.0324 - val_accuracy: 0.9250 - val_loss: 0.2047
Epoch 95/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9955 - loss: 0.0208 - val_accuracy: 0.8875 - val_loss: 0.2223
Epoch 96/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9841 - loss: 0.0363 - val_accuracy: 0.9125 - val_loss: 0.1951
Epoch 97/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9835 - loss: 0.0384 - val_accuracy: 0.9250 - val_loss: 0.1983
Epoch 98/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9801 - loss: 0.0662 - val_accuracy: 0.9375 - val_loss: 0.2212
Epoch 99/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9957 - loss: 0.0206 - val_accuracy: 0.9125 - val_loss: 0.2114
Epoch 100/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9947 - loss: 0.0318 - val_accuracy: 0.9125 - val_loss: 0.1936
Epoch 101/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0153 - val_accuracy: 0.9250 - val_loss: 0.1731
Epoch 102/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9946 - loss: 0.0219 - val_accuracy: 0.9250 - val_loss: 0.1804
Epoch 103/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 1.0000 - loss: 0.0092 - val_accuracy: 0.9125 - val_loss: 0.1641
Epoch 104/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 45ms/step - accuracy: 0.9811 - loss: 0.0325 - val_accuracy: 0.9250 - val_loss: 0.1796
Epoch 105/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9850 - loss: 0.0276 - val_accuracy: 0.9375 - val_loss: 0.1738
Epoch 106/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0074 - val_accuracy: 0.9125 - val_loss: 0.1991
Epoch 107/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9873 - loss: 0.0487 - val_accuracy: 0.9125 - val_loss: 0.1900
Epoch 108/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 45ms/step - accuracy: 0.9951 - loss: 0.0224 - val_accuracy: 0.9000 - val_loss: 0.1935
Epoch 109/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9790 - loss: 0.0544 - val_accuracy: 0.9375 - val_loss: 0.1995
Epoch 110/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0061 - val_accuracy: 0.9375 - val_loss: 0.1956
Epoch 111/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9968 - loss: 0.0158 - val_accuracy: 0.9375 - val_loss: 0.1800
Epoch 112/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9912 - loss: 0.0273 - val_accuracy: 0.9125 - val_loss: 0.1894
Epoch 113/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9939 - loss: 0.0118 - val_accuracy: 0.9250 - val_loss: 0.1858
Epoch 114/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9943 - loss: 0.0308 - val_accuracy: 0.9250 - val_loss: 0.1713
Epoch 115/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9950 - loss: 0.0152 - val_accuracy: 0.9250 - val_loss: 0.1794
Epoch 116/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0084 - val_accuracy: 0.9375 - val_loss: 0.1895
Epoch 117/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9947 - loss: 0.0174 - val_accuracy: 0.9500 - val_loss: 0.1563
Epoch 118/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 1.0000 - loss: 0.0055 - val_accuracy: 0.9500 - val_loss: 0.1477
Epoch 119/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9763 - loss: 0.0478 - val_accuracy: 0.9000 - val_loss: 0.1918
Epoch 120/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9958 - loss: 0.0135 - val_accuracy: 0.8875 - val_loss: 0.2846
Epoch 121/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9934 - loss: 0.0334 - val_accuracy: 0.9375 - val_loss: 0.1980
Epoch 122/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9943 - loss: 0.0203 - val_accuracy: 0.9500 - val_loss: 0.1832
Epoch 123/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9801 - loss: 0.0573 - val_accuracy: 0.9250 - val_loss: 0.2416
Epoch 124/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9949 - loss: 0.0334 - val_accuracy: 0.9375 - val_loss: 0.1865
Epoch 125/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9933 - loss: 0.0120 - val_accuracy: 0.9500 - val_loss: 0.1340
Epoch 126/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9944 - loss: 0.0126 - val_accuracy: 0.9250 - val_loss: 0.1565
Epoch 127/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 45ms/step - accuracy: 0.9949 - loss: 0.0143 - val_accuracy: 0.9125 - val_loss: 0.2242
Epoch 128/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9941 - loss: 0.0138 - val_accuracy: 0.9500 - val_loss: 0.1581
Epoch 129/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.9992 - loss: 0.0128 - val_accuracy: 0.9500 - val_loss: 0.1274
Epoch 130/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9966 - loss: 0.0123 - val_accuracy: 0.9625 - val_loss: 0.1514
Epoch 131/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9873 - loss: 0.0401 - val_accuracy: 0.9375 - val_loss: 0.1517
Epoch 132/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9784 - loss: 0.0407 - val_accuracy: 0.9375 - val_loss: 0.1771
Epoch 133/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9982 - loss: 0.0108 - val_accuracy: 0.9250 - val_loss: 0.2291
Epoch 134/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9957 - loss: 0.0185 - val_accuracy: 0.9000 - val_loss: 0.3030
Epoch 135/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9771 - loss: 0.0511 - val_accuracy: 0.9250 - val_loss: 0.2313
Epoch 136/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9965 - loss: 0.0162 - val_accuracy: 0.9375 - val_loss: 0.1983
Epoch 137/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9829 - loss: 0.0797 - val_accuracy: 0.9500 - val_loss: 0.1685
Epoch 138/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9910 - loss: 0.0352 - val_accuracy: 0.9625 - val_loss: 0.1578
Epoch 139/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9818 - loss: 0.0346 - val_accuracy: 0.9375 - val_loss: 0.1616
Epoch 140/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0079 - val_accuracy: 0.9375 - val_loss: 0.1702
Epoch 141/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0095 - val_accuracy: 0.9750 - val_loss: 0.1386
Epoch 142/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9987 - loss: 0.0081 - val_accuracy: 0.9750 - val_loss: 0.1187
Epoch 143/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0020 - val_accuracy: 0.9750 - val_loss: 0.1209
Epoch 144/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 49ms/step - accuracy: 0.9763 - loss: 0.0806 - val_accuracy: 0.9625 - val_loss: 0.1177
Epoch 145/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9905 - loss: 0.0263 - val_accuracy: 0.9125 - val_loss: 0.2067
Epoch 146/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0086 - val_accuracy: 0.9125 - val_loss: 0.2563
Epoch 147/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9746 - loss: 0.1065 - val_accuracy: 0.9375 - val_loss: 0.2253
Epoch 148/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9799 - loss: 0.0885 - val_accuracy: 0.9625 - val_loss: 0.1564
Epoch 149/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9955 - loss: 0.0290 - val_accuracy: 0.9250 - val_loss: 0.2414
Epoch 150/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9727 - loss: 0.0846 - val_accuracy: 0.9125 - val_loss: 0.2415
Epoch 151/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9973 - loss: 0.0157 - val_accuracy: 0.9000 - val_loss: 0.3168
Epoch 152/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9827 - loss: 0.0280 - val_accuracy: 0.9125 - val_loss: 0.2191
Epoch 153/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9856 - loss: 0.0289 - val_accuracy: 0.9500 - val_loss: 0.1684
Epoch 154/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9993 - loss: 0.0128 - val_accuracy: 0.9625 - val_loss: 0.1246
Epoch 155/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9918 - loss: 0.0194 - val_accuracy: 0.9625 - val_loss: 0.0904
Epoch 156/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 48ms/step - accuracy: 0.9992 - loss: 0.0125 - val_accuracy: 0.9625 - val_loss: 0.0854
Epoch 157/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9986 - loss: 0.0083 - val_accuracy: 0.9500 - val_loss: 0.0979
Epoch 158/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0062 - val_accuracy: 0.9625 - val_loss: 0.1077
Epoch 159/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9949 - loss: 0.0305 - val_accuracy: 0.9625 - val_loss: 0.1058
Epoch 160/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9976 - loss: 0.0084 - val_accuracy: 0.9625 - val_loss: 0.1202
Epoch 161/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0030 - val_accuracy: 0.9625 - val_loss: 0.1031
Epoch 162/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9714 - loss: 0.0519 - val_accuracy: 0.9625 - val_loss: 0.1832
Epoch 163/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0016 - val_accuracy: 0.9250 - val_loss: 0.2786
Epoch 164/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 45ms/step - accuracy: 0.9733 - loss: 0.0312 - val_accuracy: 0.8750 - val_loss: 0.2878
Epoch 165/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9897 - loss: 0.0452 - val_accuracy: 0.9375 - val_loss: 0.1482
Epoch 166/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9956 - loss: 0.0164 - val_accuracy: 0.9500 - val_loss: 0.1278
Epoch 167/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9934 - loss: 0.0399 - val_accuracy: 0.9375 - val_loss: 0.2300
Epoch 168/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9900 - loss: 0.0420 - val_accuracy: 0.8875 - val_loss: 0.5143
Epoch 169/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9869 - loss: 0.0500 - val_accuracy: 0.9125 - val_loss: 0.2374
Epoch 170/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9849 - loss: 0.0366 - val_accuracy: 0.9125 - val_loss: 0.3109
Epoch 171/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9918 - loss: 0.0244 - val_accuracy: 0.8875 - val_loss: 0.2994
Epoch 172/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9979 - loss: 0.0061 - val_accuracy: 0.9375 - val_loss: 0.2885
Epoch 173/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0073 - val_accuracy: 0.9375 - val_loss: 0.3030
Epoch 174/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9795 - loss: 0.0277 - val_accuracy: 0.8750 - val_loss: 0.4379
Epoch 175/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9966 - loss: 0.0176 - val_accuracy: 0.8750 - val_loss: 0.3758
Epoch 176/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9973 - loss: 0.0046 - val_accuracy: 0.9375 - val_loss: 0.2478
Epoch 177/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0043 - val_accuracy: 0.9375 - val_loss: 0.2529
Epoch 178/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0041 - val_accuracy: 0.9250 - val_loss: 0.2604
Epoch 179/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9973 - loss: 0.0068 - val_accuracy: 0.8875 - val_loss: 0.2902
Epoch 180/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9866 - loss: 0.0297 - val_accuracy: 0.8625 - val_loss: 0.3225
Epoch 181/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9935 - loss: 0.0085 - val_accuracy: 0.9000 - val_loss: 0.3310
Epoch 182/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9930 - loss: 0.0230 - val_accuracy: 0.8875 - val_loss: 0.4211
Epoch 183/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9981 - loss: 0.0054 - val_accuracy: 0.9125 - val_loss: 0.2929
Epoch 184/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0136 - val_accuracy: 0.9375 - val_loss: 0.2564
Epoch 185/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9907 - loss: 0.0160 - val_accuracy: 0.9000 - val_loss: 0.2726
Epoch 186/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9992 - loss: 0.0036 - val_accuracy: 0.9000 - val_loss: 0.2530
Epoch 187/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0051 - val_accuracy: 0.9250 - val_loss: 0.2283
Epoch 188/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0036 - val_accuracy: 0.9250 - val_loss: 0.2084
Epoch 189/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0012 - val_accuracy: 0.9250 - val_loss: 0.2196
Epoch 190/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0090 - val_accuracy: 0.9375 - val_loss: 0.2332
Epoch 191/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9981 - loss: 0.0096 - val_accuracy: 0.9250 - val_loss: 0.2485
Epoch 192/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9878 - loss: 0.0368 - val_accuracy: 0.9125 - val_loss: 0.3140
Epoch 193/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0013 - val_accuracy: 0.9125 - val_loss: 0.3289
Epoch 194/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0091 - val_accuracy: 0.9125 - val_loss: 0.3065
Epoch 195/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9947 - loss: 0.0131 - val_accuracy: 0.9125 - val_loss: 0.2800
Epoch 196/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9928 - loss: 0.0078 - val_accuracy: 0.9125 - val_loss: 0.2394
Epoch 197/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9957 - loss: 0.0133 - val_accuracy: 0.9000 - val_loss: 0.2319
Epoch 198/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 0.9993 - loss: 0.0031 - val_accuracy: 0.9125 - val_loss: 0.2119
Epoch 199/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0014 - val_accuracy: 0.9375 - val_loss: 0.2095
Epoch 200/200
15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 46ms/step - accuracy: 1.0000 - loss: 0.0042 - val_accuracy: 0.9375 - val_loss: 0.1972

Plot Training History

epochs_range = range(EPOCHS)

plt.figure(figsize=(14, 5))
plt.subplot(1, 2, 1)
plt.plot(
    epochs_range,
    history_model1d.history["accuracy"],
    label="Training Accuracy,1D model with non-trainable STFT",
)
plt.plot(
    epochs_range,
    history_model1d.history["val_accuracy"],
    label="Validation Accuracy, 1D model with non-trainable STFT",
)
plt.plot(
    epochs_range,
    history_model2d.history["accuracy"],
    label="Training Accuracy, 2D model with trainable STFT",
)
plt.plot(
    epochs_range,
    history_model2d.history["val_accuracy"],
    label="Validation Accuracy, 2D model with trainable STFT",
)
plt.legend(loc="lower right")
plt.title("Training and Validation Accuracy")

plt.subplot(1, 2, 2)
plt.plot(
    epochs_range,
    history_model1d.history["loss"],
    label="Training Loss,1D model with non-trainable STFT",
)
plt.plot(
    epochs_range,
    history_model1d.history["val_loss"],
    label="Validation Loss, 1D model with non-trainable STFT",
)
plt.plot(
    epochs_range,
    history_model2d.history["loss"],
    label="Training Loss, 2D model with trainable STFT",
)
plt.plot(
    epochs_range,
    history_model2d.history["val_loss"],
    label="Validation Loss, 2D model with trainable STFT",
)
plt.legend(loc="upper right")
plt.title("Training and Validation Loss")
plt.show()

png

Evaluate on Test Data

Running the models on the test set.

_, test_acc = model1d.evaluate(test_x, test_y)
print(f"1D model wit non-trainable STFT -> Test Accuracy: {test_acc * 100:.2f}%")

3/3 ━━━━━━━━━━━━━━━━━━━━ 3s 307ms/step - accuracy: 0.8148 - loss: 0.6244
1D model wit non-trainable STFT -> Test Accuracy: 82.50%

_, test_acc = model2d.evaluate(test_x, test_y)
print(f"2D model with trainable STFT -> Test Accuracy: {test_acc * 100:.2f}%")

3/3 ━━━━━━━━━━━━━━━━━━━━ 17s 546ms/step - accuracy: 0.9195 - loss: 0.5271
2D model with trainable STFT -> Test Accuracy: 92.50%
Audio Classification with the STFTSpectrogram layer
◆ Introduction
◆ Setup
Importing the necessary libraries
Define some variables
◆ Download and Preprocess the ESC-10 Dataset
Download and Extract the dataset
Read the CSV file
Define functions to read and preprocess the WAV files
Define functions to construct a TF Dataset
Create the datasets
◆ Training the Models
Create the 1D model
Create the 2D model
Plot Training History
Evaluate on Test Data