csl-final-project/model.py at main · verityw/csl-final-project · GitHub

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
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from tensorflow.keras.models import Sequential
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.callbacks import ModelCheckpoint
from tensorflow.keras.layers import Lambda, Conv2D, MaxPooling2D, Dropout, Dense, Flatten, TimeDistributed, RNN, SimpleRNN, LSTM, InputLayer
from tensorflow.keras import Input
from utils import INPUT_SHAPE, batch_generator, SEQUENCE_LEN
import argparse
import os

np.random.seed(0)

def build_pilotnet(args):
    """
    Modified NVIDIA model
    """
    model = Sequential()
    model.add(Lambda(lambda x: x/127.5-1.0, input_shape=INPUT_SHAPE))
    model.add(Conv2D(24, 5, 2, activation='elu'))
    model.add(Conv2D(36, 5, 2, activation='elu'))
    model.add(Conv2D(48, 5, 2, activation='elu'))
    model.add(Conv2D(64, 3, 1, activation='elu'))
    model.add(Conv2D(64, 3, 1, activation='elu'))
    model.add(Dropout(args.keep_prob))
    model.add(Flatten())
    model.add(Dense(100, activation='elu'))
    model.add(Dense(50, activation='elu'))
    model.add(Dense(10, activation='elu'))
    model.add(Dense(1))

    return model

def build_smallconvnet(args):
    """
    Smaller convolutional network
    """
    model = Sequential()
    model.add(Lambda(lambda x: x/127.5-1.0, input_shape=INPUT_SHAPE))
    # model.add(Conv2D(16, 5, 3, activation='elu'))
    # model.add(Conv2D(24, 5, 3, activation='elu'))
    # model.add(Conv2D(32, 5, 3, activation='elu'))
    # model.add(Conv2D(32, 3, 2, activation='elu'))
    model.add(Conv2D(16, 5, 3, activation='elu'))
    model.add(Conv2D(16, 5, 2, activation='elu'))
    model.add(Dropout(args.keep_prob))
    model.add(Flatten())
    model.add(Dense(8, activation='elu'))
    model.add(Dense(1))

    return model


def build_lstm(args):
    "Build long short-term memory RNN"
    input_shape = (16, 160, 320, 3)
    print((SEQUENCE_LEN, *INPUT_SHAPE))
    model = Sequential()
    model.add(InputLayer(input_shape=(SEQUENCE_LEN, *INPUT_SHAPE)))
    model.add(TimeDistributed(Lambda(lambda x: x/127.5-1.0)))
    model.add(TimeDistributed(Conv2D(16, 5, 3, activation='elu')))
    model.add(TimeDistributed(Conv2D(16, 5, 2, activation='elu')))
    model.add(TimeDistributed(Flatten()))
    model.add(TimeDistributed(Dropout(args.keep_prob)))
    model.add(TimeDistributed(Dense(units=16,   activation='relu')))
    model.add(LSTM(8, return_sequences=True))
    model.add(TimeDistributed(Dense(1)))
    # model.build((None, ) + input_shape)

    # model.summary()
    return model

def build_rnn(args):
    "Build RNN (fully connected recurrency)"
    input_shape = (16, 160, 320, 3)

    model = Sequential()
    model.add(InputLayer(input_shape=(SEQUENCE_LEN, *INPUT_SHAPE)))
    model.add(TimeDistributed(Lambda(lambda x: x/127.5-1.0)))
    model.add(TimeDistributed(Conv2D(16, 5, 3, activation='elu')))
    model.add(TimeDistributed(Conv2D(16, 5, 2, activation='elu')))
    model.add(TimeDistributed(Flatten()))
    model.add(TimeDistributed(Dropout(args.keep_prob)))
    model.add(TimeDistributed(Dense(units=16,   activation='relu')))
    model.add(SimpleRNN(8, return_sequences=True))
    model.add(TimeDistributed(Dense(1)))
    # model.build((None, ) + input_shape)

    # model.summary()
    return model