-
Notifications
You must be signed in to change notification settings - Fork 9
/
config.py
211 lines (189 loc) · 7.26 KB
/
config.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
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
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
'''
Server & Model configuration parameters
'''
# from models import Model_1, Model_2, Model_3, Model_4, Model_5
#### Set these:
EXPERIMENT_NUM = 2
MACHINE_NUM = 1 # 0 - mac, 1 - Titan, 2 - MARCC, 3 - MCEH
NUM_WORKERS = 4 # Cores
####
#### For SYSU:
SPLIT_NUMBER = 0
####
##################################################
# Cache directories
### Mac
if MACHINE_NUM == 0:
CACHE_METADATA = '/Users/mpeven/Projects/Activity_Recognition/cache/metadata.pickle'
CACHE_3D_OP_FLOW = '/Users/mpeven/Projects/Activity_Recognition/cache/optical_flow_3D'
CACHE_RGB_VID = '/Users/mpeven/Projects/Activity_Recognition/cache/nturgb+d_rgb'
CACHE_IR_VID = '/Users/mpeven/Projects/Activity_Recognition/cache'
CACHE_DEPTH = '/Users/mpeven/Projects/Activity_Recognition/cache'
CACHE_MASKED_DEPTH = '/Users/mpeven/Projects/Activity_Recognition/cache/nturgb+d_depth_masked'
CACHE_SKELETONS = '/Users/mpeven/Projects/Activity_Recognition/cache'
SYSU_LOCATION = '/Users/mpeven/Downloads/SYSU3DAction'
### Titan
if MACHINE_NUM == 1:
CACHE_METADATA = '/home/mike/Documents/Activity_Recognition/cache/metadata.pickle'
CACHE_RGB_VID = '/hdd/Datasets/NTU/nturgb+d_rgb'
CACHE_2D_IMAGES = '/hdd/Datasets/NTU/nturgb+d_rgb_masked'
CACHE_3D_IMAGES = '/hdd/Datasets/NTU/ntu_3D_voxel_images'
CACHE_2D_OP_FLOW = '/hdd/Datasets/NTU/nturgb+d_op_flow_2D_npy'
CACHE_2D_OP_FLOW_PNG = '/hdd/Datasets/NTU/nturgb+d_op_flow_2D_png'
CACHE_IR_VID = '/hdd/Datasets/NTU/nturgb+d_ir'
CACHE_DEPTH = '/hdd/Datasets/NTU/nturgb+d_depth'
CACHE_MASKED_DEPTH = '/hdd/Datasets/NTU/nturgb+d_depth_masked'
CACHE_SKELETONS = '/hdd/Datasets/NTU/nturgb+d_skeletons'
CACHE_3D_OP_FLOW = '/hdd/Datasets/NTU/nturgb+d_op_flow_3D'
CACHE_FEATURES_VOX_FLOW = '/home/mike/Documents/Activity_Recognition/nturgb+d_features_small'
SYSU_LOCATION = '/home/mike/Documents/SYSU'
CACHE_3D_VOX_FLOW_SYSU = '/home/mike/Documents/Activity_Recognition/SYSU_voxel_flow_3D_54'
CACHE_2D_IMAGES_SYSU = '/home/mike/Documents/Activity_Recognition/SYSU_rgb_images_5_npy'
### MARCC
if MACHINE_NUM == 2:
CACHE_METADATA = '/home-3/mpeven1@jhu.edu/work/dev_mp/nturgb_cache/metadata.pickle'
CACHE_RGB_VID = '/home-3/mpeven1@jhu.edu/data/nturgb+d_rgb'
CACHE_IR_VID = '/home-3/mpeven1@jhu.edu/data/nturgb+d_ir'
CACHE_DEPTH = '/home-3/mpeven1@jhu.edu/data/nturgb+d_depth'
CACHE_MASKED_DEPTH = '/home-3/mpeven1@jhu.edu/data/nturgb+d_depth_masked'
CACHE_SKELETONS = '/home-3/mpeven1@jhu.edu/data/nturgb+d_skeletons'
CACHE_3D_OP_FLOW = '/home-3/mpeven1@jhu.edu/work/dev_mp/nturgb_cache/optical_flow_3D'
### GPU Server
if MACHINE_NUM == 3:
CACHE_METADATA = '/home/mpeven1/rambo/home/mpeven/ntu_rgb/cache/metadata.pickle2'
CACHE_RGB_VID = '/home/mpeven1/rambo/edata/nturgb/nturgb+d_rgb'
CACHE_IR_VID = '/home/mpeven1/rambo/edata/nturgb/nturgb+d_ir'
CACHE_DEPTH = '/home/mpeven1/rambo/edata/nturgb/nturgb+d_depth'
CACHE_MASKED_DEPTH = '/home/mpeven1/rambo/edata/nturgb/nturgb+d_depth_masked'
CACHE_SKELETONS = '/home/mpeven1/rambo/edata/nturgb/nturgb+d_skeletons'
CACHE_FEATURES_VOX_FLOW = '/home/mpeven1/rambo/edata/nturgb/nturgb+d_features_small'
##################################################
# Experiments
### Default
NUM_EPOCHS = 50
HIDDEN_DIM_SIZE = 256
LSTM_DROPOUT = 0
EXPERIMENTS = [
{ ### 1 ###
'images': True,
'batch_size': 32,
}, { ### 2 ###
'images': True,
'batch_size': 4,
'cross_view': True,
}, { ### 3 ###
'images_3D': True,
'batch_size': 8,
}, { ### 4 ###
'images_3D': True,
'batch_size': 8,
'cross_view': True,
}, { ### 5 ###
'op_flow': True,
'batch_size': 8,
}, { ### 6 ###
'op_flow': True,
'batch_size': 8,
'cross_view': True,
}, { ### 7 ###
'op_flow': True,
'batch_size': 8,
'augmentation': False,
}, { ### 8 ###
'op_flow': True,
'batch_size': 8,
'augmentation': False,
'cross_view': True,
}, { ### 9 ###
'op_flow_2D': True,
'batch_size': 24,
}, { ### 10 ###
'op_flow_2D': True,
'batch_size': 16,
'cross_view': True,
}, { ### 11 ###
'op_flow_2D': True,
'batch_size': 128,
}, { ### 12 ###
'op_flow_2D': True,
'batch_size': 128,
'cross_view': True,
}, { ### 13 ###
'op_flow': True,
'batch_size': 8,
'dataset': 'SYSU',
'dataset_classes': 12,
}, { ### 14 ###
'images': True,
'batch_size': 32,
'dataset': 'SYSU',
'dataset_classes': 12,
},
]
# Set defaults
for experiment in EXPERIMENTS:
if 'dataset' not in experiment:
experiment['dataset'] = 'NTU'
experiment['dataset_classes'] = 60
if 'images' not in experiment:
experiment['images'] = False
if 'images_3D' not in experiment:
experiment['images_3D'] = False
if 'op_flow' not in experiment:
experiment['op_flow'] = False
if 'op_flow_2D' not in experiment:
experiment['op_flow_2D'] = False
if 'augmentation' not in experiment:
experiment['augmentation'] = True
if 'cross_view' not in experiment:
experiment['cross_view'] = False
if 'single_features' not in experiment:
experiment['single_feature'] = False
# Dataset config
DATASET = EXPERIMENTS[EXPERIMENT_NUM-1]['dataset']
DATASET_NUM_CLASSES = EXPERIMENTS[EXPERIMENT_NUM-1]['dataset_classes']
# Experiment config
DATA_IMAGES = EXPERIMENTS[EXPERIMENT_NUM-1]['images']
DATA_IMAGES_3D = EXPERIMENTS[EXPERIMENT_NUM-1]['images_3D']
DATA_OP_FLOW = EXPERIMENTS[EXPERIMENT_NUM-1]['op_flow']
DATA_OP_FLOW_2D = EXPERIMENTS[EXPERIMENT_NUM-1]['op_flow_2D']
DATA_AUGMENTATION = EXPERIMENTS[EXPERIMENT_NUM-1]['augmentation']
DATA_CROSS_VIEW = EXPERIMENTS[EXPERIMENT_NUM-1]['cross_view']
DATA_BATCH_SIZE = EXPERIMENTS[EXPERIMENT_NUM-1]['batch_size']
DATA_SINGLE_FEAT = EXPERIMENTS[EXPERIMENT_NUM-1]['single_feature']
def print_config():
data_desc = ""
if DATA_IMAGES:
data_desc += "2D images"
if DATA_IMAGES_3D:
data_desc += "3D images"
if DATA_OP_FLOW:
data_desc += "3D optical flow"
if DATA_OP_FLOW_2D:
data_desc += "2D optical flow"
if DATA_SINGLE_FEAT:
data_desc += " single feature"
if not DATA_AUGMENTATION:
data_desc += " (no data augmentation)"
if DATA_CROSS_VIEW:
data_desc += " - cross-view split"
else:
data_desc += " - cross-subject split"
if DATASET == "SYSU":
data_desc += " {:02}".format(SPLIT_NUMBER)
print("Experiment {:02} - {} - {}".format(EXPERIMENT_NUM, data_desc, DATASET))
from models import *
if EXPERIMENT_NUM in [1,2]:
NEURAL_NET = Model_1()
elif EXPERIMENT_NUM in [3,4]:
NEURAL_NET = Model_4()
elif EXPERIMENT_NUM in [5,6,7,8]:
NEURAL_NET = Model_2()
elif EXPERIMENT_NUM in [9,10]:
NEURAL_NET = Model_5()
elif EXPERIMENT_NUM in [11,12]:
NEURAL_NET = Model_5_small()
elif EXPERIMENT_NUM in [13]:
NEURAL_NET = Model_2()
elif EXPERIMENT_NUM in [14]:
NEURAL_NET = Model_1()