-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathstore_episodes_parallel.py
165 lines (123 loc) · 7.44 KB
/
store_episodes_parallel.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
import multiprocessing as mp
from multiprocessing import Pool, TimeoutError
import numpy as np
from datasets.dataloader import HabitatDataScene
import datasets.util.utils as utils
import os
import argparse
import torch
import random
import json
class Params(object):
def __init__(self):
self.parser = argparse.ArgumentParser()
self.parser.add_argument('--split', type=str, dest='split', default='train',
choices=['train', 'val', 'test'])
self.parser.add_argument('--grid_dim', type=int, dest='grid_dim', default=384)
self.parser.add_argument('--crop_size', type=int, dest='crop_size', default=64)
self.parser.add_argument('--cell_size', type=float, dest='cell_size', default=0.1)
self.parser.add_argument('--turn_angle', type=int, dest='turn_angle', default=30)
self.parser.add_argument('--forward_step_size', type=float, dest='forward_step_size', default=0.25)
self.parser.add_argument('--n_object_classes', type=int, dest='n_object_classes', default=27)
self.parser.add_argument('--n_spatial_classes', type=int, dest='n_spatial_classes', default=3)
self.parser.add_argument('--img_size', dest='img_size', type=int, default=256)
self.parser.add_argument('--img_segm_size', dest='img_segm_size', type=int, default=128)
self.parser.add_argument('--max_num_episodes', dest='max_num_episodes', type=int, default=2500)
self.parser.add_argument('--episode_len', type=int, dest='episode_len', default=10)
self.parser.add_argument('--truncate_ep', dest='truncate_ep', default=True,
help='truncate episode run in dataloader in order to do only the necessary steps')
self.parser.add_argument('--occ_from_depth', dest='occ_from_depth', default=True, action='store_true',
help='if enabled, uses only depth to get the ground-projected egocentric grid')
self.parser.add_argument('--scenes_list', nargs='+')
self.parser.add_argument('--root_path', type=str, dest='root_path', default="~/")
self.parser.add_argument('--episodes_path', type=str, dest='episodes_path', default="habitat-api/data/datasets/objectnav/mp3d/")
self.parser.add_argument('--ep_set', type=str, dest='ep_set', default='v1', choices=['v1','v3','v5'])
self.parser.add_argument('--episodes_root', type=str, dest='episodes_root', default="")
self.parser.add_argument('--scenes_dir', type=str, dest='scenes_dir', default='habitat-api/data/scene_datasets/')
self.parser.add_argument('--episodes_save_dir', type=str, dest='episodes_save_dir', default="mp3d_objnav_episodes_tmp/")
self.parser.add_argument('--gpu_capacity', type=int, dest='gpu_capacity', default=2)
def store_episodes(options, config_file, scene_id):
episode_save_dir = options.root_path + options.scenes_dir + options.episodes_save_dir + options.split + "/" + scene_id + "/"
if not os.path.exists(episode_save_dir):
os.makedirs(episode_save_dir)
existing_episode_list = os.listdir(episode_save_dir) # keep track of previously saved episodes
options.episodes_root = options.episodes_path + options.ep_set + '/'
data = HabitatDataScene(options, config_file, scene_id=scene_id, existing_episode_list=existing_episode_list)
print(len(data))
ep_count = len(existing_episode_list)
for i in range(len(data)):
ex = data[i]
if ep_count >= options.max_num_episodes:
break
if ex is None:
continue
ep_count+=1
scene_id = ex['scene_id']
episode_id = ex['episode_id']
abs_pose = ex['abs_pose']
ego_grid_crops_spatial = ex['ego_grid_crops_spatial'].cpu()
step_ego_grid_crops_spatial = ex['step_ego_grid_crops_spatial'].cpu()
gt_grid_crops_spatial = ex['gt_grid_crops_spatial'].cpu()
gt_grid_crops_objects = ex['gt_grid_crops_objects'].cpu()
images = ex['images'].cpu()
ssegs = ex['ssegs'].cpu()
depth_imgs = ex['depth_imgs'].cpu()
if options.truncate_ep: # assumes that the maps were created only up to the desired step
abs_pose = abs_pose[-options.episode_len:,:]
ego_grid_crops_spatial = ego_grid_crops_spatial[-options.episode_len:,:,:,:]
step_ego_grid_crops_spatial = step_ego_grid_crops_spatial[-options.episode_len:,:,:,:]
gt_grid_crops_spatial = gt_grid_crops_spatial[-options.episode_len:,:,:,:]
gt_grid_crops_objects = gt_grid_crops_objects[-options.episode_len:,:,:,:]
images = images[-options.episode_len:,:,:,:]
ssegs = ssegs[-options.episode_len:,:,:,:]
depth_imgs = depth_imgs[-options.episode_len:,:,:,:]
else: # assumes episode was run until its end
total_episode_len = ego_grid_crops_spatial.shape[0]
ind = random.randint(0, total_episode_len-options.episode_len-1)
abs_pose = abs_pose[ind:ind+options.episode_len,:]
ego_grid_crops_spatial = ego_grid_crops_spatial[ind:ind+options.episode_len,:,:,:]
step_ego_grid_crops_spatial = step_ego_grid_crops_spatial[ind:ind+options.episode_len,:,:,:]
gt_grid_crops_spatial = gt_grid_crops_spatial[ind:ind+options.episode_len,:,:,:]
gt_grid_crops_objects = gt_grid_crops_objects[ind:ind+options.episode_len,:,:,:]
images = images[ind:ind+options.episode_len,:,:,:]
ssegs = ssegs[ind:ind+options.episode_len,:,:,:]
depth_imgs = depth_imgs[ind:ind+options.episode_len,:,:,:]
print('Saving episode', ep_count, 'of id', episode_id, 'scene', scene_id)
filepath = episode_save_dir+'ep_'+str(ep_count)+'_'+str(episode_id)+"_"+scene_id
np.savez_compressed(filepath+'.npz',
abs_pose=abs_pose,
ego_grid_crops_spatial=ego_grid_crops_spatial,
step_ego_grid_crops_spatial=step_ego_grid_crops_spatial,
gt_grid_crops_spatial=gt_grid_crops_spatial,
gt_grid_crops_objects=gt_grid_crops_objects,
images=images,
ssegs=ssegs,
depth_imgs=depth_imgs
)
if __name__ == '__main__':
mp.set_start_method('forkserver', force=True)
options = Params().parser.parse_args()
print("options:")
for k in options.__dict__.keys():
print(k, options.__dict__[k])
save_path = options.root_path + options.scenes_dir + options.episodes_save_dir + options.split + "/"
if not os.path.exists(save_path):
os.makedirs(save_path)
with open(os.path.join(save_path, 'options.json'), "w") as f:
json.dump(vars(options), f, indent=4)
if options.split=="val":
config_file = "configs/my_objectnav_mp3d_val.yaml"
elif options.split=="train":
config_file = "configs/my_objectnav_mp3d_train.yaml"
else:
config_file = "configs/my_objectnav_mp3d_test.yaml"
scene_ids = options.scenes_list
# Create iterables for map function
n = len(scene_ids)
options_list = [options] * n
config_files = [config_file] * n
args = [*zip(options_list, config_files, scene_ids)]
with Pool(processes=options.gpu_capacity) as pool:
pool.starmap(store_episodes, args)
# exiting the 'with'-block has stopped the pool
print("Now the pool is closed and no longer available")