Deep learning for humans.
Keras is an API designed for human beings, not machines. Keras follows best practices for reducing cognitive load: it offers consistent & simple APIs, it minimizes the number of user actions required for common use cases, and it provides clear & actionable error messages. Keras also gives the highest priority to crafting great documentation and developer guides.
Reference :
- Documentations | Keras
https://keras.io/
Dalam project ini, kita akan membuat dan melatih sebuah model jaringan saraf tiruan sederhana untuk melakukan klasifikasi gambar menggunakan dataset MNIST, yang terdiri dari gambar-gambar angka tulisan tangan.
In this project, we will build and train a simple artificial neural network model to perform image classification using the MNIST dataset, which consists of handwritten numeric images.
Creating Directories and File Structures
❯ pwd
/Users/.../<project-name>
❯ cd <project-name>
❯ python -m venv venv
❯ source ./venv/bin/activate
❯ pip install -r requirements.txt
❯ mkdir -p data models utils
❯ touch data/__init__.py data/data_loader.py
❯ touch models/__init__.py models/neural_network.py
❯ touch utils/__init__.py utils/helper_functions.py
❯ touch train.py predict.py
❯ tree -L 2 -a -I 'README.md|.DS_Store|.git|.gitignore|venv' ./
./
├── data
│ ├── __init__.py
│ └── data_loader.py
├── models
│ ├── __init__.py
│ └── neural_network.py
├── predict.py
├── train.py
└── utils
├── __init__.py
└── helper_functions.py
3 directories, 8 files
❯ pip list Package Version ---------------------------- ----------- absl-py 2.1.0 astunparse 1.6.3 certifi 2024.2.2 charset-normalizer 3.3.2 contourpy 1.2.1 cycler 0.12.1 flatbuffers 24.3.25 fonttools 4.51.0 gast 0.5.4 google-pasta 0.2.0 grpcio 1.63.0 h5py 3.11.0 idna 3.7 keras 3.3.3 kiwisolver 1.4.5 libclang 18.1.1 Markdown 3.6 markdown-it-py 3.0.0 MarkupSafe 2.1.5 matplotlib 3.9.0 mdurl 0.1.2 ml-dtypes 0.3.2 namex 0.0.8 numpy 1.26.4 opt-einsum 3.3.0 optree 0.11.0 packaging 24.0 pillow 10.3.0 pip 22.0.4 protobuf 4.25.3 Pygments 2.18.0 pyparsing 3.1.2 python-dateutil 2.9.0.post0 requests 2.31.0 rich 13.7.1 setuptools 58.1.0 six 1.16.0 tensorboard 2.16.2 tensorboard-data-server 0.7.2 tensorflow 2.16.1 tensorflow-io-gcs-filesystem 0.37.0 termcolor 2.4.0 typing_extensions 4.11.0 urllib3 2.2.1 Werkzeug 3.0.3 wheel 0.43.0 wrapt 1.16.0
❯ vim data/data_loader.py
import numpy as np
from keras.datasets import mnist
from keras.utils import to_categorical
def load_mnist_data():
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()
train_images = train_images.reshape((60000, 28 * 28)).astype('float32') / 255
test_images = test_images.reshape((10000, 28 * 28)).astype('float32') / 255
train_labels = to_categorical(train_labels)
test_labels = to_categorical(test_labels)
return (train_images, train_labels), (test_images, test_labels)
❯ vim models/neural_network.py
from keras.models import Sequential
from keras.layers import Dense
def create_neural_network():
model = Sequential()
model.add(Dense(64, activation='relu', input_shape=(28 * 28,)))
model.add(Dense(64, activation='relu'))
model.add(Dense(10, activation='softmax'))
return model
❯ vim utils/helper_functions.py
import matplotlib.pyplot as plt
def plot_training_history(history):
acc = history.history['accuracy']
val_acc = history.history['val_accuracy']
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs = range(1, len(acc) + 1)
plt.plot(epochs, acc, 'bo', label='Training accuracy')
plt.plot(epochs, val_acc, 'b', label='Validation accuracy')
plt.title('Training and validation accuracy')
plt.legend()
plt.figure()
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
plt.show()
❯ vim train.py
from data.data_loader import load_mnist_data
from models.neural_network import create_neural_network
from utils.helper_functions import plot_training_history
(train_images, train_labels), (test_images, test_labels) = load_mnist_data()
model = create_neural_network()
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
history = model.fit(train_images, train_labels, epochs=5, batch_size=64, validation_split=0.2)
plot_training_history(history)
model.save('mnist_model.h5')
❯ vim predict.py
import numpy as np
from keras.models import load_model
from data.data_loader import load_mnist_data
(train_images, train_labels), (test_images, test_labels) = load_mnist_data()
model = load_model('mnist_model.h5')
# Prediksi label untuk beberapa gambar uji
predictions = model.predict(test_images[:5])
# Tampilkan prediksi dan label sebenarnya
for i, prediction in enumerate(predictions):
predicted_label = np.argmax(prediction)
actual_label = np.argmax(test_labels[i])
print(f'Predicted: {predicted_label}, Actual: {actual_label}')
Run the train.py file to train the model. You can do this by typing python3 train.py in the terminal.
❯ python3 train.py
Once the training is complete, the model will be saved in the file mnist_model.h5 in the same directory.
❯ tree -L 2 -a -I 'README.md|.DS_Store|.git|.gitignore|venv|gambar-petunjuk' ./
./
├── data
│ ├── __init__.py
│ ├── __pycache__
│ └── data_loader.py
├── mnist_model.h5
├── models
│ ├── __init__.py
│ ├── __pycache__
│ └── neural_network.py
├── predict.py
├── requirements.txt
├── train.py
└── utils
├── __init__.py
├── __pycache__
└── helper_functions.py
6 directories, 10 files
❯ find ./ -type f -name "*.h5" | sed 's/[^\/]*\//| /g;s/| *\([^| ]\)/+--- \1/'
| +--- mnist_model.h5
| | | | | | | | +--- vlen_string_s390x.h5
| | | | | | | | +--- vlen_string_dset_utc.h5
| | | | | | | | +--- vlen_string_dset.h5
Run the predict.py file to make predictions with the trained model. Type python3 predict.py in the terminal.
❯ python3 predict.py
WARNING:absl:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 56ms/step
Predicted: 7, Actual: 7
Predicted: 2, Actual: 2
Predicted: 1, Actual: 1
Predicted: 0, Actual: 0
Predicted: 4, Actual: 4
