-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathevaluate_motion_vae.py
213 lines (198 loc) · 10.3 KB
/
evaluate_motion_vae.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
212
213
import models.motion_vae as vae_models
from trainer.vae_trainer import *
from utils.matrix_transformer import MatrixTransformer as mt
from utils.plot_script import plot_3d_motion, draw_pose_from_cords
import utils.paramUtil as paramUtil
from utils.utils_ import *
from options.evaluate_vae_options import *
from dataProcessing import dataset
from torch.utils.data import DataLoader
if __name__ == "__main__":
parser = TestOptions()
opt = parser.parse()
joints_num = 0
input_size = 72
data = None
label_dec = None
dim_category = 31
enumerator = None
device = torch.device("cuda:" + str(opt.gpu_id) if opt.gpu_id else "cpu")
opt.save_root = os.path.join(opt.checkpoints_dir, opt.dataset_type, opt.name)
opt.model_path = os.path.join(opt.save_root, 'model')
opt.joints_path = os.path.join(opt.save_root, 'joints')
model_file_path = os.path.join(opt.model_path, opt.which_epoch + '.tar')
result_path = os.path.join(opt.result_path, opt.dataset_type, opt.name + opt.name_ext)
if opt.dataset_type == "humanact12":
dataset_path = "./dataset/humanact12"
input_size = 72
joints_num = 24
label_dec = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]
raw_offsets = paramUtil.shihao_raw_offsets
kinematic_chain = paramUtil.shihao_kinematic_chain
enumerator = paramUtil.shihao_coarse_action_enumerator
elif opt.dataset_type == "shihao":
dataset_path = "./dataset/pose"
pkl_path = './dataset/pose_shihao_merge'
input_size = 72
joints_num = 24
label_dec = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]
raw_offsets = paramUtil.shihao_raw_offsets
kinematic_chain = paramUtil.shihao_kinematic_chain
enumerator = paramUtil.shihao_coarse_action_enumerator
elif opt.dataset_type == "ntu_rgbd":
file_prefix = "./dataset/"
motion_desc_file = "motionlist.txt"
joints_num = 25
input_size = 75
labels = paramUtil.ntu_action_labels
elif opt.dataset_type == "ntu_rgbd_v2":
file_prefix = "./dataset/"
motion_desc_file = "motionlistv2.txt"
joints_num = 19
input_size = 57
labels = paramUtil.ntu_action_labels
elif opt.dataset_type == "mocap":
dataset_path = "./dataset/mocap/mocap_3djoints/"
clip_path = './dataset/mocap/pose_clip.csv'
input_size = 60
joints_num = 20
raw_offsets = paramUtil.mocap_raw_offsets
kinematic_chain = paramUtil.mocap_kinematic_chain
label_dec = [0, 1, 2, 3, 4, 5, 6, 7]
enumerator = paramUtil.mocap_action_enumerator
elif opt.dataset_type == "ntu_rgbd_vibe":
file_prefix = "./dataset"
motion_desc_file = "ntu_vibe_list.txt"
joints_num = 18
input_size = 54
label_dec = [6, 7, 8, 9, 22, 23, 24, 38, 80, 93, 99, 100, 102]
labels = paramUtil.ntu_action_labels
enumerator = paramUtil.ntu_action_enumerator
raw_offsets = paramUtil.vibe_raw_offsets
kinematic_chain = paramUtil.vibe_kinematic_chain
else:
raise NotImplementedError('This dataset is unregonized!!!')
opt.dim_category = len(label_dec)
opt.pose_dim = input_size
if opt.time_counter:
opt.input_size = input_size + opt.dim_category + 1
else:
opt.input_size = input_size + opt.dim_category
opt.output_size = input_size
model = torch.load(model_file_path)
prior_net = vae_models.GaussianGRU(opt.input_size, opt.dim_z, opt.hidden_size,
opt.prior_hidden_layers, opt.num_samples, device)
if opt.use_lie:
decoder = vae_models.DecoderGRULie(opt.input_size + opt.dim_z, opt.output_size, opt.hidden_size,
opt.decoder_hidden_layers,
opt.num_samples, device)
else:
decoder = vae_models.DecoderGRU(opt.input_size + opt.dim_z, opt.output_size, opt.hidden_size,
opt.decoder_hidden_layers,
opt.num_samples, device)
prior_net.load_state_dict(model['prior_net'])
decoder.load_state_dict(model['decoder'])
prior_net.to(device)
decoder.to(device)
if opt.use_lie:
if opt.dataset_type=='shihao':
data = dataset.MotionFolderDatasetShihaoV2(opt.clip_set, dataset_path, pkl_path, opt,
lie_enforce=opt.lie_enforce, raw_offsets=raw_offsets,
kinematic_chain=kinematic_chain)
elif opt.dataset_type == 'humanact12':
data = dataset.MotionFolderDatasetHumanAct12(dataset_path, opt, lie_enforce=opt.lie_enforce)
elif opt.dataset_type == 'ntu_rgbd_vibe':
data = dataset.MotionFolderDatasetNtuVIBE(file_prefix, motion_desc_file, labels, opt, joints_num=joints_num,
offset=True, extract_joints=paramUtil.kinect_vibe_extract_joints)
elif opt.dataset_type == 'mocap':
data = dataset.MotionFolderDatasetMocap(clip_path, dataset_path, opt)
motion_dataset = dataset.MotionDataset(data, opt)
motion_loader = DataLoader(motion_dataset, batch_size=opt.batch_size, drop_last=True, num_workers=2,
shuffle=True)
trainer = TrainerLie(motion_loader, opt, device, raw_offsets, kinematic_chain)
else:
trainer = Trainer(None, opt, device)
if opt.do_random:
fake_motion, classes = trainer.evaluate(prior_net, decoder, opt.num_samples)
fake_motion = fake_motion.cpu().numpy()
else:
categories = np.arange(opt.dim_category).repeat(opt.replic_times, axis=0)
# categories = np.arange(1).repeat(opt.replic_times, axis=0)
# categories = np.array([6]).repeat(opt.replic_times, axis=0)
num_samples = categories.shape[0]
category_oh, classes = trainer.get_cate_one_hot(categories)
fake_motion, _ = trainer.evaluate(prior_net, decoder, num_samples, category_oh)
fake_motion = fake_motion.cpu().numpy()
print(fake_motion.shape)
# print(fake_motion[:, 0, :2])
for i in range(fake_motion.shape[0]):
class_type = enumerator[label_dec[classes[i]]]
motion_orig = fake_motion[i]
if not os.path.exists(result_path):
os.makedirs(result_path)
keypoint_path = os.path.join(result_path, 'keypoint')
if not os.path.exists(keypoint_path):
os.makedirs(keypoint_path)
file_name = os.path.join(result_path, class_type + str(i) + ".gif")
offset = np.matlib.repmat(np.array([motion_orig[0, 0], motion_orig[0, 1], motion_orig[0, 2]]),
motion_orig.shape[0], joints_num)
# offset = np.tile(motion_orig[:, :3], 24)
# print(offset[1])
motion_mat = motion_orig - offset
motion_mat = motion_mat.reshape(-1, joints_num, 3)
# motion_mat[:, :, 2] *= -1
np.save(os.path.join(keypoint_path, class_type + str(i) + '_3d.npy'), motion_mat)
if opt.dataset_type == "shihao" or opt.dataset_type == "humanact12":
pose_tree = paramUtil.smpl_tree
# offset = np.tile(np.array([-0.43391575, 0.31606525, 2.57938163]), (motion_orig.shape[0], 24, 1))
# motion_3d = offset + motion_mat
# motion_3d = motion_3d.reshape(-1, joints_num, 3)
# print(motion_3d[..., 2].min(), motion_3d[..., 2].max())
# motion_2d_mat = np.zeros(motion_3d.shape[:-1] + (2,))
# motion_2d_imgs = []
# for k in range(motion_3d.shape[0]):
# motion_2d_mat[k] = project_3d_to_2d(motion_3d[k])
#
# crop_bbox, crop_size = compute_videocrop_bbox(motion_2d_mat, 100, 1.28, (1920, 1080), thresold=0)
# motion_2d_mat = crop_and_resize_motion(motion_2d_mat, crop_bbox, crop_size, (256, 200), joints_num=24)
# for k in range(motion_2d_mat.shape[0]):
# img_2d = draw_pose_from_cords((200, 256), motion_2d_mat[k], pose_tree, 2)
# motion_2d_imgs.append(img_2d)
# file_prefix = result_path + class_type
# compose_gif_img_list(motion_2d_imgs, file_prefix + '_2d.gif', duration=0.5)
# np.save(os.path.join(keypoint_path, class_type + '_3d.npy'), motion_3d)
# np.save(os.path.join(keypoint_path, class_type + '_2d.npy'), motion_2d_mat)
'''
motion_mat = mt.swap_yz(motion_mat)
motion_mat[:, :, 2] *= -1
plot_3d_motion(motion_mat, pose_tree, class_type, file_name, interval=150)
'''
ground_trajec = motion_mat[:, 0, :]
plot_3d_motion_with_trajec(motion_mat, kinematic_chain, save_path=file_name, interval=80, trajec2=ground_trajec)
elif opt.dataset_type == "ntu_rgbd":
motion_mat = mt.swap_xz(motion_mat)
pose_tree = paramUtil.kinect_tree_exclude
exclued_points = paramUtil.excluded_joint_ids
plot_3d_motion(motion_mat, pose_tree, class_type, file_name, interval=150,
excluded_joints=exclued_points)
elif opt.dataset_type == "ntu_rgbd_v2":
motion_mat = mt.swap_yz(motion_mat)
pose_tree = paramUtil.kinect_tree_v2
# exclued_points = paramUtil.excluded_joint_ids
plot_3d_motion(motion_mat, pose_tree, class_type, file_name, interval=150)
elif opt.dataset_type == "ntu_rgbd_vibe":
'''
motion_mat = mt.swap_yz(motion_mat)
motion_mat[:, :, 2] = -1 * motion_mat[:, :, 2]
motion_mat = mt.rotate_along_z(motion_mat, 180)
pose_tree = paramUtil.kinect_tree_vibe
# exclued_points = paramUtil.excluded_joint_ids
plot_3d_motion(motion_mat, pose_tree, class_type, file_name, interval=150)
'''
plot_3d_motion_v2(motion_mat, kinematic_chain, save_path=file_name, interval=80)
elif opt.dataset_type == "mocap":
pose_tree = paramUtil.kinect_tree_mocap
# plot_3d_motion(motion_mat, pose_tree, class_type, file_name, interval=150)
ground_trajec = motion_mat[:, 0, :]
plot_3d_motion_with_trajec(motion_mat, kinematic_chain, save_path=file_name, interval=80,
trajec1=ground_trajec, dataset="mocap")