-
Notifications
You must be signed in to change notification settings - Fork 5
/
vis.py
197 lines (170 loc) · 8.11 KB
/
vis.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
from loader import build_loader
from model import build_model
from trainer import build_trainer
from inference import build_inference, Inference_COCO
from utils import ensure_dir
import logging, coloredlogs
import argparse
import yaml
import os
import torch
import random
import json
from os import listdir
from os.path import isfile, join
import pickle
from tqdm import tqdm
# setting parser
parser = argparse.ArgumentParser()
parser.add_argument('--cfg_path', type=str, default='configs/')
parser.add_argument('--default_cfg_path', type=str, default='configs/default.yaml')
parser.add_argument('--checkpoint', type=str, default=None)
parser.add_argument('--test_output_dir', type=str, default='./images/Appendix/output_4')
parser.add_argument('--layout_output_dir', type=str, default='./images/Appendix/output_4/layout')
parser.add_argument('--test_scene_graphs_dir', type=str,
default='./images/vg_msdn_v24_sg')
parser.add_argument('--image_id', type=str, default='')
parser.add_argument('--repeat', type=int, default=1)
opt = parser.parse_args()
# setting config file
with open(opt.default_cfg_path, 'r') as f:
cfg = yaml.safe_load(f)
with open(opt.cfg_path, 'r') as f:
cfg.update(yaml.safe_load(f))
# handle dir for saving
ensure_dir(cfg['OUTPUT']['OUTPUT_DIR'])
ensure_dir(cfg['TEST']['OUTPUT_DIR'])
# setting logger
handlers = [logging.FileHandler(os.path.join(cfg['OUTPUT']['OUTPUT_DIR'],
'output_eval_from_sg.log'), mode = 'w'),
logging.StreamHandler()]
logging.basicConfig(handlers = handlers, level=logging.INFO)
logger = logging.getLogger('root')
coloredlogs.install(logger = logger, fmt='%(asctime)s [%(name)s] %(levelname)s %(message)s')
logger.info('Setup output directory - {}.'.format(cfg['OUTPUT']['OUTPUT_DIR']))
if __name__ == '__main__':
model = build_model(cfg)
assert opt.checkpoint is not None, 'Please provide model ckpt for testing'
checkpoint = torch.load(opt.checkpoint)
model.load_state_dict(checkpoint['state_dict'])
ly_save_path = opt.layout_output_dir
data_dir = cfg['DATASETS']['DATA_DIR_PATH']
vocab_dic_path = os.path.join(data_dir, 'object_pred_idx_to_name.pkl')
with open(vocab_dic_path, 'rb') as file:
vocab_dict = pickle.load(file)
infer = Inference_COCO(save_dir = opt.test_output_dir, vocab_dict = vocab_dict)
# Load the scene graphs
onlyfiles = [f for f in listdir(opt.test_scene_graphs_dir) if \
isfile(join(opt.test_scene_graphs_dir, f))]
if opt.image_id != '':
iter_idx = onlyfiles.index(opt.image_id+'.json')
# print(iter_idx)
for dd in range(opt.repeat):
for idx, filename in tqdm(enumerate(onlyfiles)):
if opt.image_id != '' and iter_idx != idx:
continue
file_path = os.path.join(opt.test_scene_graphs_dir, filename)
with open(file_path, 'r') as f:
scene_graphs = json.load(f)
# print(scene_graphs)
input_dict = dict()
input_dict['image_id'] = str(scene_graphs['image_id'])+'_{}'.format(dd)
input_dict['dataset_idx'] = scene_graphs['dataset_idx']
input_dict['tensor_list'] = []
print(scene_graphs['objects'])
print(scene_graphs['relationships'])
# random suffle
for jj in range(3):
num_objs = len(scene_graphs['objects'])
sampling = random.choices([i for i in range(num_objs)], k=2)
sour_idx, targ_idx = sampling[0], sampling[1]
sour_obj = scene_graphs['objects'][sour_idx]
targ_obj = scene_graphs['objects'][targ_idx]
scene_graphs['objects'][targ_idx] = sour_obj
scene_graphs['objects'][sour_idx] = targ_obj
for i in range(len(scene_graphs['relationships'])):
if scene_graphs['relationships'][i][0] == sour_idx:
scene_graphs['relationships'][i][0] = targ_idx
elif scene_graphs['relationships'][i][0] == targ_idx:
scene_graphs['relationships'][i][0] = sour_idx
if scene_graphs['relationships'][i][2] == sour_idx:
scene_graphs['relationships'][i][2] = targ_idx
elif scene_graphs['relationships'][i][2] == targ_idx:
scene_graphs['relationships'][i][2] = sour_idx
print(scene_graphs['objects'])
print(scene_graphs['relationships'])
tmp_name_to_idx = dict()
for i, (idx, name) in enumerate(infer.vocab_dict.items()):
tmp_name_to_idx[name] = idx
# tmp_name_to_idx['None'] = 2
# [1, 128]
# create obj_class_id list
class_ids_list = []
for obj_name in scene_graphs['objects']:
class_ids_list.append(tmp_name_to_idx[obj_name])
obj_ids_list = []
for i in class_ids_list:
obj_ids_list.append(0)
# create input_token
input_token = torch.zeros(128).type(torch.int64)
input_obj_id = torch.zeros(128).type(torch.int64)
segment_label = torch.zeros(128).type(torch.int64)
token_type = torch.zeros(128).type(torch.int64)
input_token[0] = 1
segment_label[0] = 1
sent_p = 1
random.shuffle(scene_graphs['relationships'])
for i in range(len(scene_graphs['relationships'])):
pair = scene_graphs['relationships'][i]
input_token[sent_p] = class_ids_list[pair[0]]
input_token[sent_p+1] = tmp_name_to_idx[pair[1]]
input_token[sent_p+2] = class_ids_list[pair[2]]
input_token[sent_p+3] = 2
# save obj_id
obj_ids_list[pair[0]] = pair[0] + 1
obj_ids_list[pair[2]] = pair[2] + 1
if class_ids_list[pair[0]] == 3: obj_ids_list[pair[0]] = 0
if class_ids_list[pair[2]] == 3: obj_ids_list[pair[2]] = 0
input_obj_id[sent_p] = pair[0] + 1
input_obj_id[sent_p+1] = 0
input_obj_id[sent_p+2] = pair[2] + 1
input_obj_id[sent_p+3] = 0
if class_ids_list[pair[0]] == 3: input_obj_id[sent_p] = 0
if class_ids_list[pair[2]] == 3: input_obj_id[sent_p+2] = 0
segment_label[sent_p] = int((sent_p - 1) / 4 + 1)
segment_label[sent_p+1] = int((sent_p - 1) / 4 + 1)
segment_label[sent_p+2] = int((sent_p - 1) / 4 + 1)
segment_label[sent_p+3] = int((sent_p - 1) / 4 + 1)
token_type[sent_p] = 1
token_type[sent_p+1] = 2
token_type[sent_p+2] = 3
sent_p += 4
# __image__
__image__ = tmp_name_to_idx['__image__']
__in_image__ = tmp_name_to_idx['__in_image__']
for i in range(len(class_ids_list)):
O = class_ids_list[i]
input_token[sent_p] = O
input_token[sent_p+1] = __in_image__
input_token[sent_p+2] = __image__
input_token[sent_p+3] = 2
input_obj_id[sent_p] = obj_ids_list[i]
input_obj_id[sent_p+1] = 0
input_obj_id[sent_p+2] = 0
input_obj_id[sent_p+3] = 0
segment_label[sent_p] = int((sent_p - 1) / 4 + 1)
segment_label[sent_p+1] = int((sent_p - 1) / 4 + 1)
segment_label[sent_p+2] = int((sent_p - 1) / 4 + 1)
segment_label[sent_p+3] = int((sent_p - 1) / 4 + 1)
token_type[sent_p] = 1
token_type[sent_p+1] = 2
token_type[sent_p+2] = 3
sent_p += 4
tmp_list = []
tmp_list.append(input_token)
tmp_list.append(input_obj_id)
tmp_list.append(segment_label)
tmp_list.append(token_type)
# print(tmp_list[0])
input_dict['tensor_list'] = tmp_list
infer.check_from_sg(input_dict, model, layout_save=ly_save_path)