-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathtest.py
161 lines (136 loc) · 7.9 KB
/
test.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
import os
import argparse
import torch
import numpy as np
from tqdm import tqdm
import data_loader.data_loaders as module_data
import model.model as module_arch
from parse_config import ConfigParser
from model.metric import chamfer_distance_naive, surfaceSampling, IOU, fscore
from model.loss import PI_value_generator, PI_funcs_generator
from utils.util import gen_polynomial_orders
import matplotlib.pyplot as plt
def main(config):
logger = config.get_logger('test')
# setup data_loader instances
test_data_loader = config.init_obj('test_data_loader', module_data)
# build model architecture
model = config.init_obj('arch', module_arch)
logger.info(model)
logger.info('Loading checkpoint: {} ...'.format(config.resume))
checkpoint = torch.load(config.resume)
state_dict = checkpoint['state_dict']
if config['n_gpu'] > 1:
model = torch.nn.DataParallel(model)
model.load_state_dict(state_dict)
checkname = str(config.run_id)
os.makedirs('./visualization/'+checkname+"/functions/", exist_ok=True)
os.makedirs('./visualization/'+checkname+"/valid_indices/", exist_ok=True)
cls1_name = ['02691156', '02828884', '02933112', '02958343', '03001627', '03211117', '03636649', '03691459', '04090263', '04256520', '04379243', '04401088', '04530566']
for i in range(len(cls1_name)):
os.makedirs('./visualization/'+checkname+'/functions/'+cls1_name[i], exist_ok=True)
os.makedirs('./visualization/'+checkname+'/valid_indices/'+cls1_name[i], exist_ok=True)
# prepare model for testing
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = model.to(device)
model.eval()
sum_iou_cls = torch.zeros(13).to('cuda')
sum_f_cls = torch.zeros(13).to('cuda')
sum_chamfer_cls = torch.zeros(13).to('cuda')
sum_iou_cls_mean = torch.zeros(13).to('cuda')
sum_f_cls_mean = torch.zeros(13).to('cuda')
sum_cham_cls_mean = torch.zeros(13).to('cuda')
count_cls = torch.zeros(13).to('cuda')
with torch.no_grad():
polyorder_cpu = gen_polynomial_orders(4)
for img_H, target in tqdm(test_data_loader):
img_H = img_H.to('cuda')
for key in target:
if target[key].size==0: print(target['directory']); raise
if key!='directory':
target[key] = target[key].to('cuda')
polycoeff, _, A_10x10 = model(img_H)
polyorder = torch.from_numpy(gen_polynomial_orders(4)).to('cuda')
PI_funcs_inout = PI_funcs_generator(target['inoutpts'], polycoeff, polyorder)
PI_value_inout, _ = PI_value_generator(torch.tanh(PI_funcs_inout))
batchiou = IOU(PI_value_inout, target)[:]
batchf = fscore(PI_value_inout, target)[:]
sampled_point = surfaceSampling(polycoeff, test_data_loader.dataset.allpoints) # on points
batchchamferL1 = chamfer_distance_naive(sampled_point, target['onpts'])
surfs, validinds = coeff2polystr(polycoeff.detach().cpu().numpy(), polyorder_cpu, A_10x10, PI_funcs_inout)
for ii in range(batchiou.shape[0]):
sum_iou_cls_mean[target['class_num'][ii]] += batchiou[ii]
sum_f_cls_mean[target['class_num'][ii]] += batchf[ii]
sum_cham_cls_mean[target['class_num'][ii]] += batchchamferL1[ii]
if batchiou[ii] >= sum_iou_cls[target['class_num'][ii]]:
print('Check: ',batchiou[ii], target['directory'][ii])
sum_iou_cls[target['class_num'][ii]] = max(batchiou[ii],sum_iou_cls[target['class_num'][ii]])
sum_f_cls[target['class_num'][ii]] = max(batchiou[ii],sum_f_cls[target['class_num'][ii]])
sum_chamfer_cls[target['class_num'][ii]] = max(batchiou[ii],sum_chamfer_cls[target['class_num'][ii]])
count_cls[target['class_num'][ii]] += 1
# save sample images, or do something with output here
with open('./visualization/'+checkname+'/functions/'+'0'+str(target['directory'][ii][0].item())+'/'+str(target['directory'][ii][1].item())+'.txt', 'w') as f:
f.write(surfs[ii])
with open('./visualization/'+checkname+'/valid_indices/'+'0'+str(target['directory'][ii][0].item())+'/'+str(target['directory'][ii][1].item())+'.txt', 'w') as f:
f.write(validinds[ii])
cls_name = ['plane', 'bench', 'cabinet', 'car', 'chair', 'display', 'lamp', 'speaker', 'rifle', 'sofa', 'table', 'phone', 'vessel']
iou_per_cls = sum_iou_cls_mean/count_cls
f_per_cls = sum_f_cls_mean/count_cls
cham_per_cls = sum_cham_cls_mean/count_cls
print("class names:", cls_name)
print("# samples :", count_cls)
for ii in range(13):
print(cls_name[ii]+" IoU: {}".format(iou_per_cls[ii]))
print(cls_name[ii]+" chamfer: {}".format(cham_per_cls[ii]))
print(cls_name[ii]+" F: {}".format(f_per_cls[ii]))
print("#####################")
print("naive average IoU: {}".format(torch.mean(iou_per_cls).item()))
print("Total average IoU: {}".format(torch.sum(sum_iou_cls_mean)/torch.sum(count_cls)))
print("naive average F: {}".format(torch.mean(f_per_cls).item()))
print("Total average F: {}".format(torch.sum(sum_f_cls_mean)/torch.sum(count_cls)))
print("naive average chamfer: {}".format(torch.mean(cham_per_cls).item()))
print("Total average chamfer: {}".format(torch.sum(sum_cham_cls_mean)/torch.sum(count_cls)))
def coeff2polystr(polycoeff, polyorder, A_10x10, PI_funcs_inout):
"""
polycoeff (=Params):
(batch, num_params, num_functions) = (batch, 35, 32)
polyorders:
(num_params, 3) = (35, 3) - degree of x,y,z for each term in polynomial.
Since we use 4th-polynomials, the number of parameters is determined by (4+1)(4+1+1)(4+1+2)/6 = 35
"""
batchsurfs = ["" for i in range(polycoeff.shape[0])]
batchind = ["" for i in range(polycoeff.shape[0])]
for batch in range(polycoeff.shape[0]): #batch
surface = ""
c = 0
list_valid = []
for subf in range(polycoeff.shape[2]): #100 = 25*4
f_tmp = ""
A_10x10[batch,:,:,subf] = A_10x10[batch,:,:,subf]/torch.norm(A_10x10[batch,:,:,subf])
if torch.min(torch.eig(A_10x10[batch,:,:,subf])[0][:,0])<-0.0 and torch.max(torch.eig(A_10x10[batch,:,:,subf])[0][:,0])>0 and torch.sum(torch.abs(torch.prod(torch.eig(A_10x10[batch,:,:,subf])[0][:,1])))==0 and torch.min(PI_funcs_inout[batch,:,subf])<0:
list_valid.append(subf)
for term in range(polycoeff.shape[1]): #35
f_tmp = f_tmp + '+('+str(polycoeff[batch, term, subf])+')*(x^'+str(int(polyorder[term,0]))+')*(y^'+str(int(polyorder[term,1]))+')*(z^'+str(int(polyorder[term,2]))+')'
if c==0:
surface = surface + f_tmp[1:]+'\n'
else:
surface = 'min(' + surface + ',' + f_tmp[1:]+')'+'\n'
c+=1
valids = ""
for sub_valid in list_valid:
valids = valids + str(sub_valid) + ' '
batchind[batch] = batchind[batch] + valids #+ '-' + str(R[batch,0])
batchsurfs[batch] = batchsurfs[batch] + surface #+ '-' + str(R[batch,0])
return batchsurfs, batchind
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='3DIAS_PyTorch')
parser.add_argument('-c', '--config', default=None, type=str,
help='config file path (default: None)')
parser.add_argument('-r', '--resume', default=None, type=str,
help='path to latest checkpoint (default: None)')
parser.add_argument('-d', '--device', default=None, type=str,
help='indices of GPUs to enable (default: all)')
parser.add_argument('-t', '--tag', default=None, type=str,
help='experience name in tensorboard (default: None)')
config = ConfigParser.from_args(parser.parse_args()) #, options)
main(config)