-
Notifications
You must be signed in to change notification settings - Fork 0
/
nn.txt
42 lines (37 loc) · 1.74 KB
/
nn.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
Very fast to train, good accuracy for easy problems (e.g. currency_type) but struggle with hard problems (e.g. item_links),
stop improving really fast
keras.layers.experimental.preprocessing.Rescaling(1. / 255, input_shape=(size[0], size[1], 3)),
keras.layers.Conv2D(32, (3, 3), padding='same', activation="relu", input_shape=(size[0], size[1], 3)),
keras.layers.MaxPooling2D((2, 2), strides=2),
keras.layers.Conv2D(64, (3, 3), padding='same', activation="relu"),
keras.layers.MaxPooling2D((2, 2), strides=2),
keras.layers.Flatten(),
keras.layers.Dense(100, activation="relu"),
keras.layers.Dense(num_classes, activation="softmax")
-----
Quite fast to train, very good accuracy, good improvements the more epochs are used
Conv2D(16, (3, 3), activation="relu", input_shape=(46, 46, 3)),
MaxPool2D(2, 2),
Conv2D(32, (3, 3), activation="relu"),
MaxPool2D(2, 2),
Conv2D(64, (3, 3), activation="relu"),
MaxPool2D(2, 2),
Flatten(),
Dense(512, activation="relu"),
Dense(num_classes, activation="softmax")
-----
Very slow to train (needs a lot of epoch to achieve anything), accurate with easy medium problems but require an enormous amount of epochs to solve hard problems,
not much improvements per epoch, stop improving at some point
keras.layers.experimental.preprocessing.Rescaling(1. / 255, input_shape=(size[0], size[1], 3)),
keras.layers.Conv2D(8, 3, padding='same', activation='relu'),
keras.layers.MaxPooling2D(),
keras.layers.Dropout(0.3),
keras.layers.Conv2D(16, 3, padding='same', activation='relu'),
keras.layers.MaxPooling2D(),
keras.layers.Dropout(0.3),
keras.layers.Conv2D(32, 3, padding='same', activation='relu'),
keras.layers.MaxPooling2D(),
keras.layers.Dropout(0.3),
keras.layers.Flatten(),
keras.layers.Dense(128, activation='relu'),
keras.layers.Dense(num_classes)