Setup Base Directory¶
In [ ]:
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
In [ ]:
BaseDir = '/content/drive/MyDrive/Colab Notebooks/Klasifikasi-Gambar/'
Download Library¶
In [ ]:
%%capture
!pip install --upgrade tensorflow
!pip install tensorflowjs
!pip install tensorflow
Import Library¶
In [ ]:
import os
import shutil
import numpy as np
import pandas as pd
import matplotlib
import matplotlib.pyplot as plt
import seaborn as sns
import sklearn
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, confusion_matrix, accuracy_score
from PIL import Image
import cv2
import tensorflow as tf
import tensorflowjs as tfjs
from tensorflow import keras
from tensorflow.keras import layers, models, optimizers
from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau
from tensorflow.keras.applications import MobileNetV2, mobilenet_v2
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
import warnings
tf.get_logger().setLevel('ERROR')
warnings.filterwarnings('ignore')
In [ ]:
print(f"{'Numpy Version':20}: {np.__version__}")
print(f"{'Pandas Version':20}: {pd.__version__}")
print(f"{'Tensorflow Version':20}: {tf.__version__}")
print(f"{'TensorflowJS Version':20}: {tfjs.__version__}")
print(f"{'Matplotlib Version':20}: {matplotlib.__version__}")
print(f"{'Scikit-Learn Version':20}: {sklearn.__version__}")
Numpy Version : 1.26.4 Pandas Version : 2.2.2 Tensorflow Version : 2.15.1 TensorflowJS Version: 4.20.0 Matplotlib Version : 3.7.5 Scikit-Learn Version: 1.2.2
In [ ]:
from google.colab import files
files.upload()
Data Loading¶
In [ ]:
# %%capture
!unzip fruits.zip
Merged Dataset¶
Menggabungkan gambar dari 3 folder yang berbeda.
In [ ]:
import os
import shutil
# Define the paths
training_dataset_path = '/content/fruits/Training'
test_dataset_path = '/content/fruits/Test'
validation_dataset_path = '/content/fruits/Validation'
dataset_path = BaseDir+'dataset/fruits'
def copy_contents(src, dst):
for item in os.listdir(src):
s = os.path.join(src, item)
d = os.path.join(dst, item)
if os.path.isdir(s):
if not os.path.exists(d):
os.makedirs(d)
copy_contents(s, d)
else:
shutil.copy2(s, d)
# Create the new directory for the merged dataset
os.makedirs(dataset_path, exist_ok=True)
# Copy contents into the new directory
copy_contents(training_dataset_path, dataset_path)
copy_contents(validation_dataset_path, dataset_path)
copy_contents(test_dataset_path, dataset_path)
Menampilkan kelas yang terdapat pada dataset¶
In [ ]:
# Classes
DIR = BaseDir+'dataset/fruits'
classes = [i for i in os.listdir(DIR) if '.' not in i]
classes
Out[ ]:
['cucumber_1', 'pear_3', 'apple_red_3', 'apple_golden_1', 'cabbage_white_1', 'apple_granny_smith_1', 'apple_crimson_snow_1', 'apple_golden_2', 'carrot_1', 'apple_braeburn_1', 'apple_golden_3', 'apple_red_delicios_1', 'apple_pink_lady_1', 'pear_1', 'zucchini_dark_1', 'apple_red_2', 'apple_red_yellow_1', 'eggplant_long_1', 'apple_hit_1', 'cucumber_3', 'zucchini_1', 'apple_rotten_1', 'apple_6', 'apple_red_1']
In [ ]:
# List ekstensi file gambar yang diizinkan
allowed_extensions = ['.png', '.jpg', '.jpeg']
# Inisialisasi list untuk label dan path
label = []
path = []
for dirname, _, filenames in os.walk(DIR):
for filename in filenames:
# Jika ekstensi file ada dalam allowed_extensions
if os.path.splitext(filename)[-1].lower() in allowed_extensions:
label.append(os.path.split(dirname)[-1])
path.append(os.path.join(dirname, filename))
# Create df
df = pd.DataFrame(columns=['path','label'])
df['path']=path
df['label']=label
In [ ]:
df.head()
| path | label | |
|---|---|---|
| 0 | /content/drive/MyDrive/Colab Notebooks/Klasifi... | apple_6 |
| 1 | /content/drive/MyDrive/Colab Notebooks/Klasifi... | apple_6 |
| 2 | /content/drive/MyDrive/Colab Notebooks/Klasifi... | apple_6 |
| 3 | /content/drive/MyDrive/Colab Notebooks/Klasifi... | apple_6 |
| 4 | /content/drive/MyDrive/Colab Notebooks/Klasifi... | apple_6 |
In [ ]:
df.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 12455 entries, 0 to 12454 Data columns (total 2 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 path 12455 non-null object 1 label 12455 non-null object dtypes: object(2) memory usage: 194.7+ KB
Terdapat 12.455 data gambar yang siap digunakan.
✅ Kriteria > 10.000 data gambar terpenuhi
In [ ]:
# Check value counts
df['label'].value_counts()
Out[ ]:
label apple_hit_1 936 pear_1 650 apple_granny_smith_1 640 apple_braeburn_1 640 apple_rotten_1 638 apple_crimson_snow_1 636 apple_red_2 636 apple_golden_3 632 apple_6 630 apple_pink_lady_1 625 apple_red_1 618 apple_red_yellow_1 616 apple_golden_2 616 apple_golden_1 616 apple_red_delicios_1 600 apple_red_3 562 cucumber_3 325 eggplant_long_1 320 zucchini_1 320 zucchini_dark_1 320 pear_3 287 carrot_1 201 cucumber_1 200 cabbage_white_1 191 Name: count, dtype: int64
In [ ]:
# df['label'].unique()
Cek Resolution¶
Pada langkah ini, hanya akan menampilkan 5 resolusi gambar yang berbeda. Jika menampilkan semuanya notebook dapat mengalami not responding.
In [ ]:
def print_images_resolution(directory):
unique_sizes = set()
total_images = 0
for subdir in os.listdir(directory):
subdir_path = os.path.join(directory, subdir)
image_files = os.listdir(subdir_path)
num_images = len(image_files)
print(f"{subdir}: {num_images}")
total_images += num_images
for img_file in image_files:
if len(unique_sizes) >= 5:
break
img_path = os.path.join(subdir_path, img_file)
with Image.open(img_path) as img:
unique_sizes.add(img.size)
for size in list(unique_sizes)[:5]: # Menampilkan 5 resolusi unik
print(f"- {size}")
print("---------------")
unique_sizes.clear()
print(f"\nTotal: {total_images}")
In [ ]:
print_images_resolution(DIR)
cucumber_1: 200 - (319, 800) - (313, 803) - (337, 800) - (329, 795) - (359, 791) --------------- pear_3: 287 - (623, 747) - (608, 742) - (595, 745) - (607, 739) - (603, 744) --------------- apple_red_3: 562 - (571, 563) - (601, 565) - (544, 546) - (539, 557) - (572, 564) --------------- apple_golden_1: 616 - (360, 320) - (363, 322) - (361, 367) - (353, 368) - (360, 368) --------------- cabbage_white_1: 191 - (736, 705) - (781, 768) - (740, 718) - (766, 726) - (727, 705) --------------- apple_granny_smith_1: 640 - (474, 461) - (445, 462) - (478, 462) - (508, 459) - (469, 462) --------------- apple_crimson_snow_1: 636 - (585, 569) - (647, 577) - (611, 552) - (572, 532) - (568, 570) --------------- apple_golden_2: 616 - (418, 396) - (425, 395) - (418, 398) - (406, 443) - (411, 440) --------------- carrot_1: 201 - (205, 633) - (193, 632) - (184, 652) - (182, 630) - (178, 645) --------------- apple_braeburn_1: 640 - (474, 393) - (474, 457) - (467, 412) - (471, 458) - (474, 458) --------------- apple_golden_3: 632 - (438, 383) - (404, 433) - (417, 378) - (415, 436) - (398, 433) --------------- apple_red_delicios_1: 600 - (605, 553) - (538, 582) - (577, 539) - (602, 555) - (557, 591) --------------- apple_pink_lady_1: 625 - (430, 446) - (465, 422) - (473, 420) - (469, 420) - (417, 445) --------------- pear_1: 650 - (540, 435) - (415, 522) - (537, 434) - (415, 525) - (551, 433) --------------- zucchini_dark_1: 320 - (257, 924) - (312, 890) - (282, 852) - (218, 836) - (235, 849) --------------- apple_red_2: 636 - (445, 453) - (458, 450) - (434, 410) - (447, 451) - (443, 406) --------------- apple_red_yellow_1: 616 - (327, 310) - (322, 340) - (353, 340) - (335, 306) - (352, 339) --------------- eggplant_long_1: 320 - (197, 703) - (177, 699) - (123, 739) - (155, 722) - (105, 741) --------------- apple_hit_1: 936 - (780, 907) - (820, 898) - (890, 815) - (891, 793) - (848, 801) --------------- cucumber_3: 325 - (265, 917) - (242, 949) - (368, 971) - (286, 922) - (243, 928) --------------- zucchini_1: 320 - (276, 1033) - (288, 1029) - (283, 1010) - (289, 1009) - (281, 1033) --------------- apple_rotten_1: 638 - (335, 317) - (339, 342) - (353, 344) - (350, 317) - (342, 340) --------------- apple_6: 630 - (322, 253) - (314, 262) - (314, 320) - (310, 320) - (313, 320) --------------- apple_red_1: 618 - (424, 470) - (458, 412) - (421, 446) - (460, 410) - (446, 414) --------------- Total: 12455
Show Images Sample¶
In [ ]:
Terlihat dari gambar yang ditampilkan memiliki resolusi/ ukuran yang berbeda.
Split Dataset¶
In [ ]:
# Train Test Split
train_df, test_df = train_test_split(df, test_size=0.2, shuffle=True, random_state=42)
In [ ]:
# Check shape of df
print(train_df.shape)
print(test_df.shape)
(9964, 2) (2491, 2)
Data Preprocessing¶
In [ ]:
train_generator = ImageDataGenerator(
preprocessing_function=preprocess_input
)
test_generator = ImageDataGenerator(
preprocessing_function=preprocess_input
)
In [ ]:
train_images = train_generator.flow_from_dataframe(
dataframe=train_df,
x_col='path',
y_col='label',
target_size=(224, 224),
color_mode='rgb',
# color_mode='grayscale',
class_mode='categorical',
batch_size=32,
shuffle=False,
seed=42
)
test_images = test_generator.flow_from_dataframe(
dataframe=test_df,
x_col='path',
y_col='label',
target_size=(224, 224),
color_mode='rgb',
# color_mode='grayscale',
class_mode='categorical',
batch_size=32,
shuffle=False
)
Found 9964 validated image filenames belonging to 24 classes. Found 2491 validated image filenames belonging to 24 classes.
Plotting images after pre-processing¶
In [ ]:
Modelling¶
Sequential¶
In [ ]:
model = tf.keras.Sequential([
tf.keras.layers.Conv2D(32, (3, 3), input_shape=(224, 224, 3)),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.LeakyReLU(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(64, (3, 3)),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.LeakyReLU(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Conv2D(128, (3, 3)),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.LeakyReLU(),
tf.keras.layers.MaxPooling2D((2, 2)),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(128),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.LeakyReLU(),
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(24, activation='softmax')
])
model.summary()
Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_1 (Conv2D) (None, 222, 222, 32) 896
batch_normalization (Batch (None, 222, 222, 32) 128
Normalization)
leaky_re_lu (LeakyReLU) (None, 222, 222, 32) 0
max_pooling2d_1 (MaxPoolin (None, 111, 111, 32) 0
g2D)
conv2d_2 (Conv2D) (None, 109, 109, 64) 18496
batch_normalization_1 (Bat (None, 109, 109, 64) 256
chNormalization)
leaky_re_lu_1 (LeakyReLU) (None, 109, 109, 64) 0
max_pooling2d_2 (MaxPoolin (None, 54, 54, 64) 0
g2D)
conv2d_3 (Conv2D) (None, 52, 52, 128) 73856
batch_normalization_2 (Bat (None, 52, 52, 128) 512
chNormalization)
leaky_re_lu_2 (LeakyReLU) (None, 52, 52, 128) 0
max_pooling2d_3 (MaxPoolin (None, 26, 26, 128) 0
g2D)
flatten_1 (Flatten) (None, 86528) 0
dense_3 (Dense) (None, 128) 11075712
batch_normalization_3 (Bat (None, 128) 512
chNormalization)
leaky_re_lu_3 (LeakyReLU) (None, 128) 0
dropout (Dropout) (None, 128) 0
dense_4 (Dense) (None, 24) 3096
=================================================================
Total params: 11173464 (42.62 MB)
Trainable params: 11172760 (42.62 MB)
Non-trainable params: 704 (2.75 KB)
_________________________________________________________________
In [ ]:
# Compile the model
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
# Define callbacks
early_stopping = EarlyStopping(monitor='val_loss', patience=3, restore_best_weights=True)
reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2, patience=2, min_lr=0.001)
# Train the model
history = model.fit(train_images,
validation_data=test_images,
epochs=10,
callbacks=[early_stopping, reduce_lr])
Epoch 1/10 312/312 [==============================] - 44s 122ms/step - loss: 0.2747 - accuracy: 0.9386 - val_loss: 0.6819 - val_accuracy: 0.7896 - lr: 0.0010 Epoch 2/10 312/312 [==============================] - 31s 101ms/step - loss: 0.0167 - accuracy: 0.9993 - val_loss: 0.0036 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 3/10 312/312 [==============================] - 32s 102ms/step - loss: 0.0319 - accuracy: 0.9940 - val_loss: 0.0105 - val_accuracy: 0.9972 - lr: 0.0010 Epoch 4/10 312/312 [==============================] - 32s 102ms/step - loss: 0.0063 - accuracy: 0.9997 - val_loss: 4.3834e-04 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 5/10 312/312 [==============================] - 32s 104ms/step - loss: 0.0037 - accuracy: 0.9997 - val_loss: 0.0040 - val_accuracy: 0.9980 - lr: 0.0010 Epoch 6/10 312/312 [==============================] - 32s 101ms/step - loss: 0.0031 - accuracy: 0.9999 - val_loss: 2.8158e-04 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 7/10 312/312 [==============================] - 32s 102ms/step - loss: 0.0014 - accuracy: 1.0000 - val_loss: 6.3109e-05 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 8/10 312/312 [==============================] - 36s 114ms/step - loss: 0.0010 - accuracy: 1.0000 - val_loss: 2.8823e-05 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 9/10 312/312 [==============================] - 33s 104ms/step - loss: 6.2821e-04 - accuracy: 1.0000 - val_loss: 1.8058e-05 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 10/10 312/312 [==============================] - 33s 105ms/step - loss: 6.6366e-04 - accuracy: 1.0000 - val_loss: 2.2270e-05 - val_accuracy: 1.0000 - lr: 0.0010
Plot Accuracy and Validation¶
In [ ]:
# Plotting Akurasi
plt.figure(figsize=(12, 6))
plt.subplot(1, 2, 1)
plt.plot(history.history['accuracy'], label='Train Accuracy')
plt.plot(history.history['val_accuracy'], label='Validation Accuracy')
plt.title('Model Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(loc='lower right')
# Plotting Loss
plt.subplot(1, 2, 2)
plt.plot(history.history['loss'], label='Train Loss')
plt.plot(history.history['val_loss'], label='Validation Loss')
plt.title('Model Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend(loc='upper right')
plt.show()
Test Accuracy¶
In [ ]:
results = model.evaluate(test_images, verbose=0)
print("Test Loss: {:.5f}".format(results[0]))
print("Test Accuracy: {:.2f}%".format(results[1] * 100))
Test Loss: 0.00002 Test Accuracy: 100.00%
MobileNetV2¶
In [ ]:
# Load pretrained MobileNetV2 model without the top layers
pretrained_model = tf.keras.applications.MobileNetV2(
input_shape=(224, 224, 3),
include_top=False,
weights='imagenet'
)
pretrained_model.trainable = False
# Create a new Sequential model
model2 = models.Sequential()
# Add the pretrained model
model2.add(pretrained_model)
# Add Conv2D layer
model2.add(layers.Conv2D(32, (3, 3), activation='relu'))
# Add Pooling layer
model2.add(layers.MaxPooling2D((2, 2)))
# Flatten the output from the pretrained model
model2.add(layers.Flatten())
# Add Dense layers
model2.add(layers.Dense(128, activation='relu'))
model2.add(layers.Dense(128, activation='relu'))
# Add the output layer
model2.add(layers.Dense(24, activation='softmax'))
# Display the model summary
model2.summary()
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_224_no_top.h5
9406464/9406464 [==============================] - 0s 0us/step
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
mobilenetv2_1.00_224 (Func (None, 7, 7, 1280) 2257984
tional)
conv2d (Conv2D) (None, 5, 5, 32) 368672
max_pooling2d (MaxPooling2 (None, 2, 2, 32) 0
D)
flatten (Flatten) (None, 128) 0
dense (Dense) (None, 128) 16512
dense_1 (Dense) (None, 128) 16512
dense_2 (Dense) (None, 24) 3096
=================================================================
Total params: 2662776 (10.16 MB)
Trainable params: 404792 (1.54 MB)
Non-trainable params: 2257984 (8.61 MB)
_________________________________________________________________
In [ ]:
model2.compile(optimizer='adam', loss='categorical_crossentropy',metrics=['accuracy'])
# Define callbacks
early_stopping = EarlyStopping(monitor='val_loss', patience=4, restore_best_weights=True)
reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2, patience=2, min_lr=0.0001)
history = model2.fit(train_images,
validation_data=test_images,
epochs=5,
callbacks=[early_stopping, reduce_lr]
)
Epoch 1/5 312/312 [==============================] - 38s 110ms/step - loss: 0.0302 - accuracy: 0.9909 - val_loss: 1.3443e-04 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 2/5 312/312 [==============================] - 33s 104ms/step - loss: 4.8600e-05 - accuracy: 1.0000 - val_loss: 4.0772e-05 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 3/5 312/312 [==============================] - 32s 103ms/step - loss: 1.9244e-05 - accuracy: 1.0000 - val_loss: 2.1851e-05 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 4/5 312/312 [==============================] - 33s 104ms/step - loss: 9.1450e-06 - accuracy: 1.0000 - val_loss: 1.0300e-05 - val_accuracy: 1.0000 - lr: 0.0010 Epoch 5/5 312/312 [==============================] - 32s 102ms/step - loss: 4.1812e-06 - accuracy: 1.0000 - val_loss: 5.3353e-06 - val_accuracy: 1.0000 - lr: 0.0010
Plot Accuracy and Validation¶
In [ ]:
# Plotting Akurasi
plt.figure(figsize=(12, 6))
plt.subplot(1, 2, 1)
plt.plot(history.history['accuracy'], label='Train Accuracy')
plt.plot(history.history['val_accuracy'], label='Validation Accuracy')
plt.title('Model Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(loc='lower right')
# Plotting Loss
plt.subplot(1, 2, 2)
plt.plot(history.history['loss'], label='Train Loss')
plt.plot(history.history['val_loss'], label='Validation Loss')
plt.title('Model Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend(loc='upper right')
plt.show()
Save Model¶
Saved Model¶
In [ ]:
os.makedirs(BaseDir+'saved_model', exist_ok=True)
tf.saved_model.save(model, 'saved_model')
TFLITE¶
In [ ]:
# Convert the model to TF-Lite format
converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()
# Save the TF-Lite model to a specific folder
os.makedirs(BaseDir+'tf_lite', exist_ok=True)
folder_path = BaseDir+'tf_lite'
file_path = f'{folder_path}/model.tflite'
with open(file_path, 'wb') as f:
f.write(tflite_model)
In [ ]:
def recreate_labels():
labels = [folder for folder in os.listdir(DIR) if not folder.startswith('.')]
with open(BaseDir+'tf_lite/labels.txt', 'w') as file:
for label in labels:
file.write(label)
file.write('\n')
recreate_labels()
TFJS¶
In [ ]:
import tensorflowjs as tfjs
os.makedirs(BaseDir+'tfjs_model', exist_ok=True)
tfjs.converters.save_keras_model(model, BaseDir+'tfjs_model')
Inference¶
In [ ]:
In [ ]:
image_path = BaseDir+'dataset/fruits/cucumber_1/r0_49.jpg'
input_data = preprocess_image(image_path)
interpreter.set_tensor(input_details[0]['index'], input_data)
interpreter.invoke()
output_data = interpreter.get_tensor(output_details[0]['index'])
predicted_class_idx = np.argmax(output_data, axis=1)[0]
predicted_class = labels[predicted_class_idx]
print('Predicted Class:', predicted_class)
Predicted Class: cucumber_1
In [ ]:
image_path = BaseDir+'dataset/fruits/pear_1/r1_292.jpg'
input_data = preprocess_image(image_path)
interpreter.set_tensor(input_details[0]['index'], input_data)
interpreter.invoke()
output_data = interpreter.get_tensor(output_details[0]['index'])
predicted_class_idx = np.argmax(output_data, axis=1)[0]
predicted_class = labels[predicted_class_idx]
print('Predicted Class:', predicted_class)
Predicted Class: pear_1
Conclusion¶
Berdasarkan percobaan, dihasilkan untuk:
- Model 1
- Model : Sequential
- Akurasi : 100 %
- Validasi akurasi : 100 %
- Model 2
- Model : Sequential dengan pretrained MobileNetV2
- Akurasi : 100 %
- Validasi akurasi : 100 %
Sehingga model tersebut dapat digunakan untuk memprediksi gambar buah (fruits) dimasa mendatang.
- Model 1
Hasil model mampu memprediksi gambar buah (fruits) yang diinputkan melalui inference.