-
Notifications
You must be signed in to change notification settings - Fork 5
/
Constants.py
65 lines (54 loc) · 3.03 KB
/
Constants.py
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
#Image File Location
IMAGE_FILE_LOCATION = './images/'
PR_SAVE_LOCATION = "./object_detection_test/"
SMALL_IMAGES_LOCATION = "./small_images/"
#Small Ship Data File Names
FULL_SMALL_IMAGE_DATASET = './smallShipData.json'
TEST_SMALL_IMAGE_DATASET = 'small_test_data.txt'
TRAINING_SMALL_IMAGE_DATASET = 'small_training_data.txt'
TRAINED_CNN_MODEL = 'trained_CNN.h5'
#New Small Ship Data File Names
OTHER_TEST_SMALL_IMAGE_DATASET = 'other_small_test_data.json'
OTHER_TRAINING_SMALL_IMAGE_DATASET = 'other_small_training_data.json'
OTHER_TRAINED_CNN_MODEL = 'other_trained_CNN.h5'
#Same as version4 (This one sucked)
OTHER_GENERATOR_TRAINED_CNN_MODEL = 'other_generated_trained_CNN.h5'
#Models after data augmentation
#Unless otherwise specified, learing rate = 00.1, momentum = 1e-6, decay = 0.9
#Default value I saw on website with 50 samples
GENERATOR_TRAINED_CNN_MODEL = 'generated_trained_CNN.h5'
#Same but with 100 samples
GENERATOR_TRAINED_CNN_MODEL_TWO = 'generated_version2_trained_CNN.h5'
#Adding rotational augmentation
GENERATOR_TRAINED_CNN_MODEL_THREE = 'generated_version3_trained_CNN.h5'
#Adding featurewise_center (Not very good)
GENERATOR_TRAINED_CNN_MODEL_FOUR = 'generated_version4_trained_CNN.h5'
#Added 2 fully connected layers 100/100
GENERATOR_TRAINED_CNN_MODEL_FIVE = 'generated_version5_trained_CNN.h5'
#Same as version 5 but 200/100 (Best for the final object detection ship predictions)
GENERATOR_TRAINED_CNN_MODEL_SIX = 'generated_version6_trained_CNN.h5'
#Models with transfer learning
# Learning rate = 0.001, decay = 0.9, epoch = 10
# Training Values loss: 0.0160 - binary_accuracy: 0.9990
# Validation Values Loss: 0.1450024089826009 Accuracy: 0.9860000014305115
PRETRAINED_CNN_MODEL_ONE = "pretrained_attempt_1.h5"
#Learning rate = 0.0001, decay = 0.75, epoch = 10
# Training Values loss: 0.0042 - binary_accuracy: 0.9995
# Validation Values Loss: 0.06005835090763867 Accuracy: 0.977999997138977
PRETRAINED_CNN_MODEL_TWO = "pretrained_attempt_2.h5"
# Learning rate = 0.00001, decay = 0.65, epoch = 30
# Training Values loss: 0.0239 - binary_accuracy: 0.9945
# Validation Values Loss: 0.1274039216339588 Accuracy: 0.9580000042915344
PRETRAINED_CNN_MODEL_THREE = "pretrained_attempt_3.h5"
# Learning rate = 0.001, decay = 0.9, epoch = 10 with data aumentation
# Training Values loss: 4.1746 - binary_accuracy: 0.7410
# Validation Values val_loss: 3.7394 - val_binary_accuracy: 0.7680
PRETRAINED_CNN_MODEL_FOUR = "pretrained_attempt_4.h5"
#Learning rate = 0.0001, decay = 0.75, epoch = 10 with data augmentation
# Training Values loss: 0.0434 - binary_accuracy: 0.9853
# Validation Values val_loss: 0.0395 - val_binary_accuracy: 0.9860
PRETRAINED_CNN_MODEL_FIVE = "pretrained_attempt_5.h5"
# Learning rate = 0.00001, decay = 0.65, epoch = 30 with data augmentation
# Training Values loss: 0.0779 - binary_accuracy: 0.9732
# Validation Values val_loss: 0.0575 - val_binary_accuracy: 0.9700
PRETRAINED_CNN_MODEL_SIX = "pretrained_attempt_6.h5"