-
Notifications
You must be signed in to change notification settings - Fork 0
/
config.example.yml
139 lines (122 loc) · 4.02 KB
/
config.example.yml
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
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
run:
# save / load configuration
resume_file:
# folder to store run data
run_folder: /path/to/runfolder
phase : defs.phase.train
# phase : defs.phase.val
# data parameters; #a list of input data
data:
dataset1-train:
data_path: /path/to/pathlabels.txt
prepend_folder: ""
raw_image_shape: (240, 320, 3)
image_shape: (227, 227, 3)
mean_image: [99.197148 ,105.293620 ,109.503945 ]
data_format: defs.data_format.tfrecord
frame_format: "jpg"
imgproc: [defs.imgproc.rand_crop, defs.imgproc.rand_mirror, defs.imgproc.sub_mean]
batch_item: defs.batch_item.default
phase: defs.phase.train
tag: defs.dataset_tag.main
dataset1-test:
data_path: /media/npittaras/SAMSUNG/Data/msc-thesis/ccv/serialized/frames_test/test.existing.txt
prepend_folder: ""
raw_image_shape: (240, 320, 3)
image_shape: (227, 227, 3)
mean_image: [99.197148 ,105.293620 ,109.503945 ]
data_format: defs.data_format.tfrecord
frame_format: "jpg"
imgproc: [defs.imgproc.center_crop, defs.imgproc.sub_mean]
batch_item: defs.batch_item.default
phase: defs.phase.val
tag: defs.dataset_tag.main
network:
num_classes : 20
pipelines:
- frames:
input: defs.dataset_tag.main
representation: defs.representation.dcnn
load_weights: models/alexnet/bvlc_alexnet.npy
frame_encoding_layer: "fc7"
#input_shape: (227, 227,3)
- spectros:
input: defs.dataset_tag.aux
representation: defs.representation.dcnn
load_weights: models/alexnet/bvlc_alexnet.npy
frame_encoding_layer: "fc7"
#input_shape: (227, 227,3)
frame_fusion: [defs.fusion_type.early, defs.fusion_method.avg]
- ibias:
input: [frames, spectros]
representation: defs.representation.nop
classifier: defs.classifier.lstm
lstm_params: [500, 2, defs.fusion_method.avg, defs.combo.ibias]
# phase settings
train:
# training settings
batch_size: 20
epochs : 10
optimizer : defs.optim.sgd
#momentum: 0.9
base_lr: 0.05
lr_mult : None
lr_decay : [ defs.decay.exp, defs.periodicity.drops, 100, 0.96 ]
#clip_grads : (-1.,1.)
clip_grads : None
clip_norm : 10
#clip_grads : None
dropout_keep_prob : 0.5
val:
# validation settings
batch_size: 2
logits_save_interval : -1
clip_fusion : [defs.fusion_type.late, defs.fusion_method.avg]
logging:
# logging
save_freq_per_epoch : 1
level : logging.DEBUG
print_tensors : False
tensorboard_folder : "tensorboard"
# only gmail support
email_notify: [[email protected], [email protected]]
# imgdesc vars
captioning:
caption_search : defs.caption_search.max
eval_type : defs.eval_type.coco
caption_ground_truth : "path/to/caption/gt"
word_embeddings_file : "path/to/embeddings"
serialize:
# path to prepend to each image path
path_prepend_folder: "path/to/images/folder"
# video / image list to serialize
input_files: [ "/path/to/imgfile1", "/path/to/imgfile2"]
# run type
do_shuffle: True
do_serialize: True
do_validate: True
validate_pcnt: 10
num_threads: 2
num_items_per_thread: 20
run_id: None
output_folder: "path/to/output/folder/"
generation_error: defs.generation_error.compromise
# video frames generation parameters
clip_offset_or_num: 2
num_frames_per_clip: 16
raw_image_shape: (240,320,3)
clipframe_mode: defs.clipframe_mode.iterative
frame_format: "jpg"
captions:
# vocabulary to encode captions or produce embeddings.Set to None to generate.
vocabulary_file: None
# caption files to encode or to generate vocabulary
caption_files: ["/path/to/captionfile1","/path/to/captionfile2"]
caption_file_formats: ["coco","flickr"]
vocab_replacement_file: None
word_count_thresh: 5
caption_max_length: 50
randomize_missing_embeddings: False
# embedding generation files
embeddings_file: "/path/to/embedding/matrix"
embeddings_file_type: "glove"