1. import tensorflow as tf
2. from tensorflow.keras.optimizers import RMSprop
3. from tensorflow.keras.preprocessing.image import ImageDataGenerator
4. from sklearn.model_selection import train_test_split
5. import os
6.  
7. !wget --no-check-certificate \
8.  https://dicodingacademy.blob.core.windows.net/picodiploma/ml_pemula_academy/rockpaperscissors.zip \
9.  -O /tmp/rockpaperscissors.zip
10.  
11.  
12. import zipfile
13. local_zip = '/tmp/rockpaperscissors.zip'
14. zip_ref = zipfile.ZipFile(local_zip,'r')
15. zip_ref.extractall('/tmp')
16. zip_ref.close()
17.  
18. base_dir ='/tmp/rockpaperscissors/rps-cv-images'
19.  
20. os.listdir(base_dir)
21.  
22. base_dir='/tmp/rockpaperscissors'
23. train_dir=os.path.join(base_dir,'train')
24. validation_dir=os.path.join(base_dir,'validation')
25. rock_dir=os.path.join(base_dir,'rock')
26. paper_dir=os.path.join(base_dir,'paper')
27. scissor_dir=os.path.join(base_dir,'scissors')
28.  
29.  
30. os.mkdir(train_dir)
31. os.mkdir(validation_dir)
32.  
33. train_rock=os.path.join(train_dir,'rock')
34. train_paper=os.path.join(train_dir,'paper')
35. train_scissor=os.path.join(train_dir,'scissor')
36. val_rock=os.path.join(validation_dir,'rock')
37. val_paper=os.path.join(validation_dir,'paper')
38. val_scissor=os.path.join(validation_dir,'scissor')
39.  
40. os.mkdir(train_rock)
41. os.mkdir(train_paper)
42. os.mkdir(train_scissor)
43. os.mkdir(val_rock)
44. os.mkdir(val_paper)
45. os.mkdir(val_scissor)
46.  
47. #split data rock dengan proporsi 60 40  antara train dan validation
48. train_rock_dir,val_rock_dir=train_test_split(os.listdir(rock_dir), test_size=0.4)
49.  
50. #split data paper dengan proporsi 60 40 antara train dan validation
51. train_paper_dir,val_paper_dir=train_test_split(os.listdir(paper_dir), test_size=0.4)
52.  
53. #split data scissor dengan proporsi 60 40 antara train dan validation
54. train_scissor_dir,val_scissors_dir=train_test_split(os.listdir(scissor_dir),test_size=0.4)
55.  
56. import tensorflow as tf
57. print(tf.__version__)
58.  
59. base_dir='/tmp/rockpaperscissors'
60. train_dir=os.path.join(base_dir,'train')
61. val_dir=os.path.join(base_dir,'validation')
62.  
63. os.listdir(train_dir)
64.  
65. os.listdir(val_dir)
66.  
67. #membuat directory rock pada direktori data training
68. train_rock_dir=os.path.join(train_dir,'rock')
69.  
70. #membuat directory paper pada directory data training
71. train_paper_dir=os.path.join(train_dir,'paper')
72.  
73. #membuat directory scissor pada directory data training
74. train_scissor_dir=os.path.join(train_dir,'scissors')
75.  
76. #membuat directory rock pada directory data validasi
77. val_rock_dir=os.path.join(val_dir,'rock')
78.  
79. #membuat directory paper pada directory data validasi
80. val_paper_dir=os.path.join(val_dir,'paper')
81.  
82. #membuat directory scissor pada directory data validasi
83. val_scissors_dir=os.path.join(val_dir,'scissor')
84.  
85. from tensorflow.keras.preprocessing.image import ImageDataGenerator
86.  
87. train_datagen=ImageDataGenerator(
88.                   rescale=1./255,
89.                   rotation_range=20,
90.                   horizontal_flip=True,
91.                   shear_range=0.2,
92.                   fill_mode='wrap',
93.                   validation_split=0.4)
94.  
95. test_datagen=ImageDataGenerator(
96.                   rescale=1./255,
97.                   rotation_range=20,
98.                   horizontal_flip=True,
99.                   shear_range=0.2,
100.                   fill_mode='wrap',
101.                   validation_split=0.4)
102.  
103. train_generator=train_datagen.flow_from_directory(
104.         base_dir,
105.         target_size=(150,150),
106.         class_mode='categorical',
107.         shuffle=True,
108.         subset='training'
109.         )
110.  
111. validation_generator=test_datagen.flow_from_directory(
112.         base_dir,
113.         target_size=(150,150),
114.         class_mode='categorical',
115.         shuffle=True,
116.         subset='validation',
117.         )
118.  
119. model = tf.keras.models.Sequential([
120.     tf.keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(150, 150, 3)),
121.     tf.keras.layers.MaxPooling2D(2, 2),
122.     tf.keras.layers.Conv2D(64, (3,3), activation='relu'),
123.     tf.keras.layers.MaxPooling2D(2,2),
124.     tf.keras.layers.Conv2D(128, (3,3), activation='relu'),
125.     tf.keras.layers.MaxPooling2D(2,2),
126.     tf.keras.layers.Conv2D(128, (3,3), activation='relu'),
127.     tf.keras.layers.MaxPooling2D(2,2),
128.     tf.keras.layers.Flatten(),
129.     tf.keras.layers.Dense(512, activation='relu'),
130.     tf.keras.layers.Dense(3, activation='softmax')
131. ])
132.  
133. # compile model dengan 'adam' optimizer loss function 'categorical_crossentropy'
134. model.compile(loss='categorical_crossentropy',
135.               optimizer=tf.optimizers.Adam(),
136.               metrics=['accuracy'])
137.  
138. # latih model dengan model.fit
139. model.fit(
140.       train_generator,
141.       steps_per_epoch=25,  # berapa batch yang akan dieksekusi pada setiap epoch
142.       epochs=20, # tambahkan eposchs jika akurasi model belum optimal
143.       validation_data=validation_generator, # menampilkan akurasi pengujian data validasi
144.       validation_steps=5,  # berapa batch yang akan dieksekusi pada setiap epoch
145.       verbose=2)