-
Notifications
You must be signed in to change notification settings - Fork 144
/
train_linemod.py
359 lines (302 loc) · 16.4 KB
/
train_linemod.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
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
import sys
from skimage.io import imsave
sys.path.append('.')
sys.path.append('..')
from lib.ransac_voting_gpu_layer.ransac_voting_gpu import ransac_voting_layer_v3, \
estimate_voting_distribution_with_mean, ransac_voting_layer_v5, ransac_motion_voting
from lib.networks.model_repository import *
from lib.datasets.linemod_dataset import LineModDatasetRealAug, ImageSizeBatchSampler, VotingType
from lib.utils.data_utils import LineModImageDB, OcclusionLineModImageDB, TruncatedLineModImageDB
from lib.utils.arg_utils import args
from lib.utils.draw_utils import visualize_bounding_box, imagenet_to_uint8, visualize_mask, visualize_points, img_pts_to_pts_img
from lib.utils.base_utils import save_pickle
import json
from lib.utils.evaluation_utils import Evaluator
from lib.utils.net_utils import AverageMeter, Recorder, smooth_l1_loss, \
load_model, save_model, adjust_learning_rate, compute_precision_recall, set_learning_rate
from lib.utils.config import cfg
from torch.nn import DataParallel
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler
from torch import nn, optim
import torch
import torch.nn.functional as F
import os
import time
from collections import OrderedDict
import random
import numpy as np
with open(args.cfg_file,'r') as f:
train_cfg=json.load(f)
train_cfg['model_name']='{}_{}'.format(args.linemod_cls,train_cfg['model_name'])
if train_cfg['vote_type']=='BB8C':
vote_type=VotingType.BB8C
vote_num=9
elif train_cfg['vote_type']=='BB8S':
vote_type=VotingType.BB8S
vote_num=9
elif train_cfg['vote_type']=='Farthest':
vote_type=VotingType.Farthest
vote_num=9
elif train_cfg['vote_type']=='Farthest4':
vote_type=VotingType.Farthest4
vote_num=5
elif train_cfg['vote_type']=='Farthest12':
vote_type=VotingType.Farthest12
vote_num=13
elif train_cfg['vote_type']=='Farthest16':
vote_type=VotingType.Farthest16
vote_num=17
else:
assert(train_cfg['vote_type']=='BB8')
vote_type=VotingType.BB8
vote_num=8
seg_loss_rec = AverageMeter()
ver_loss_rec = AverageMeter()
precision_rec = AverageMeter()
recall_rec = AverageMeter()
recs=[seg_loss_rec,ver_loss_rec,precision_rec,recall_rec]
recs_names=['scalar/seg','scalar/ver','scalar/precision','scalar/recall']
data_time = AverageMeter()
batch_time = AverageMeter()
recorder = Recorder(True,os.path.join(cfg.REC_DIR,train_cfg['model_name']),
os.path.join(cfg.REC_DIR,train_cfg['model_name']+'.log'))
# network_time,voting_time,load_time=[],[],[]
# poses_pr=[]
# poses_gt=[]
class NetWrapper(nn.Module):
def __init__(self,net):
super(NetWrapper,self).__init__()
self.net=net
self.criterion=nn.CrossEntropyLoss(reduce=False)
def forward(self, image, mask, vertex, vertex_weights):
seg_pred, vertex_pred = self.net(image)
loss_seg = self.criterion(seg_pred, mask)
loss_seg = torch.mean(loss_seg.view(loss_seg.shape[0],-1),1)
loss_vertex = smooth_l1_loss(vertex_pred, vertex, vertex_weights, reduce=False)
precision, recall = compute_precision_recall(seg_pred, mask)
return seg_pred, vertex_pred, loss_seg, loss_vertex, precision, recall
class EvalWrapper(nn.Module):
def forward(self, seg_pred, vertex_pred, use_argmax=True, use_uncertainty=False):
vertex_pred=vertex_pred.permute(0,2,3,1)
b,h,w,vn_2=vertex_pred.shape
vertex_pred=vertex_pred.view(b,h,w,vn_2//2,2)
if use_argmax:
mask=torch.argmax(seg_pred,1)
else:
mask=seg_pred
if use_uncertainty:
return ransac_voting_layer_v5(mask,vertex_pred,128,inlier_thresh=0.99,max_num=100)
else:
return ransac_voting_layer_v3(mask,vertex_pred,128,inlier_thresh=0.99,max_num=100)
class MotionEvalWrapper(nn.Module):
def forward(self, seg_pred, vertex_pred, use_argmax=True, use_uncertainty=False):
vertex_pred=vertex_pred.permute(0,2,3,1)
b,h,w,vn_2=vertex_pred.shape
vertex_pred=vertex_pred.view(b,h,w,vn_2//2,2)
if use_argmax:
mask=torch.argmax(seg_pred,1)
else:
mask=seg_pred
return ransac_motion_voting(mask, vertex_pred)
class UncertaintyEvalWrapper(nn.Module):
def forward(self, seg_pred, vertex_pred, use_argmax=True):
vertex_pred=vertex_pred.permute(0,2,3,1)
b,h,w,vn_2=vertex_pred.shape
vertex_pred=vertex_pred.view(b,h,w,vn_2//2,2)
if use_argmax:
mask=torch.argmax(seg_pred,1)
else:
mask=seg_pred
mean=ransac_voting_layer_v3(mask, vertex_pred, 512, inlier_thresh=0.99)
mean, var=estimate_voting_distribution_with_mean(mask,vertex_pred,mean)
return mean, var
def train(net, optimizer, dataloader, epoch):
for rec in recs: rec.reset()
data_time.reset()
batch_time.reset()
train_begin=time.time()
net.train()
size = len(dataloader)
end=time.time()
for idx, data in enumerate(dataloader):
image, mask, vertex, vertex_weights, pose, _ = [d.cuda() for d in data]
data_time.update(time.time()-end)
seg_pred, vertex_pred, loss_seg, loss_vertex, precision, recall = net(image, mask, vertex, vertex_weights)
loss_seg, loss_vertex, precision, recall=[torch.mean(val) for val in (loss_seg, loss_vertex, precision, recall)]
loss = loss_seg + loss_vertex * train_cfg['vertex_loss_ratio']
vals=(loss_seg,loss_vertex,precision,recall)
for rec,val in zip(recs,vals): rec.update(val)
optimizer.zero_grad()
loss.backward()
optimizer.step()
batch_time.update(time.time()-end)
end=time.time()
if idx % train_cfg['loss_rec_step'] == 0:
step = epoch * size + idx
losses_batch=OrderedDict()
for name,rec in zip(recs_names,recs): losses_batch['train/'+name]=rec.avg
recorder.rec_loss_batch(losses_batch,step,epoch)
for rec in recs: rec.reset()
data_time.reset()
batch_time.reset()
if idx % train_cfg['img_rec_step'] == 0:
batch_size = image.shape[0]
nrow = 5 if batch_size > 5 else batch_size
recorder.rec_segmentation(F.softmax(seg_pred, dim=1), num_classes=2, nrow=nrow, step=step, name='train/image/seg')
recorder.rec_vertex(vertex_pred, vertex_weights, nrow=4, step=step, name='train/image/ver')
print('epoch {} training cost {} s'.format(epoch,time.time()-train_begin))
def val(net, dataloader, epoch, val_prefix='val', use_camera_intrinsic=False, use_motion=False):
for rec in recs: rec.reset()
test_begin = time.time()
evaluator = Evaluator()
eval_net=DataParallel(EvalWrapper().cuda()) if not use_motion else DataParallel(MotionEvalWrapper().cuda())
uncertain_eval_net=DataParallel(UncertaintyEvalWrapper().cuda())
net.eval()
for idx, data in enumerate(dataloader):
if use_camera_intrinsic:
image, mask, vertex, vertex_weights, pose, corner_target, Ks = [d.cuda() for d in data]
else:
image, mask, vertex, vertex_weights, pose, corner_target = [d.cuda() for d in data]
with torch.no_grad():
seg_pred, vertex_pred, loss_seg, loss_vertex, precision, recall = net(image, mask, vertex, vertex_weights)
loss_seg, loss_vertex, precision, recall=[torch.mean(val) for val in (loss_seg, loss_vertex, precision, recall)]
if (train_cfg['eval_epoch']
and epoch%train_cfg['eval_inter']==0
and epoch>=train_cfg['eval_epoch_begin']) or args.test_model:
if args.use_uncertainty_pnp:
mean,cov_inv=uncertain_eval_net(seg_pred,vertex_pred)
mean=mean.cpu().numpy()
cov_inv=cov_inv.cpu().numpy()
else:
corner_pred=eval_net(seg_pred,vertex_pred).cpu().detach().numpy()
pose=pose.cpu().numpy()
b=pose.shape[0]
pose_preds=[]
for bi in range(b):
intri_type='use_intrinsic' if use_camera_intrinsic else 'linemod'
K=Ks[bi].cpu().numpy() if use_camera_intrinsic else None
if args.use_uncertainty_pnp:
pose_preds.append(evaluator.evaluate_uncertainty(mean[bi],cov_inv[bi],pose[bi],args.linemod_cls,
intri_type,vote_type,intri_matrix=K))
else:
pose_preds.append(evaluator.evaluate(corner_pred[bi],pose[bi],args.linemod_cls,intri_type,
vote_type,intri_matrix=K))
if args.save_inter_result:
mask_pr = torch.argmax(seg_pred, 1).cpu().detach().numpy()
mask_gt = mask.cpu().detach().numpy()
# assume batch size = 1
imsave(os.path.join(args.save_inter_dir, '{}_mask_pr.png'.format(idx)), mask_pr[0])
imsave(os.path.join(args.save_inter_dir, '{}_mask_gt.png'.format(idx)), mask_gt[0])
imsave(os.path.join(args.save_inter_dir, '{}_rgb.png'.format(idx)),
imagenet_to_uint8(image.cpu().detach().numpy()[0]))
save_pickle([pose_preds[0],pose[0]],os.path.join(args.save_inter_dir, '{}_pose.pkl'.format(idx)))
vals=[loss_seg,loss_vertex,precision,recall]
for rec,val in zip(recs,vals): rec.update(val)
with torch.no_grad():
batch_size = image.shape[0]
nrow = 5 if batch_size > 5 else batch_size
recorder.rec_segmentation(F.softmax(seg_pred, dim=1), num_classes=2, nrow=nrow,
step=epoch, name='{}/image/seg'.format(val_prefix))
recorder.rec_vertex(vertex_pred, vertex_weights, nrow=4, step=epoch, name='{}/image/ver'.format(val_prefix))
losses_batch=OrderedDict()
for name, rec in zip(recs_names, recs): losses_batch['{}/'.format(val_prefix) + name] = rec.avg
if (train_cfg['eval_epoch']
and epoch%train_cfg['eval_inter']==0
and epoch>=train_cfg['eval_epoch_begin']) or args.test_model:
proj_err,add,cm=evaluator.average_precision(False)
losses_batch['{}/scalar/projection_error'.format(val_prefix)]=proj_err
losses_batch['{}/scalar/add'.format(val_prefix)]=add
losses_batch['{}/scalar/cm'.format(val_prefix)]=cm
recorder.rec_loss_batch(losses_batch, epoch, epoch, val_prefix)
for rec in recs: rec.reset()
print('epoch {} {} cost {} s'.format(epoch,val_prefix,time.time()-test_begin))
def train_net():
net=Resnet18_8s(ver_dim=vote_num*2, seg_dim=2)
net=NetWrapper(net)
net=DataParallel(net).cuda()
optimizer = optim.Adam(net.parameters(), lr=train_cfg['lr'])
model_dir=os.path.join(cfg.MODEL_DIR,train_cfg['model_name'])
motion_model=train_cfg['motion_model']
print('motion state {}'.format(motion_model))
if args.test_model:
begin_epoch=load_model(net.module.net, optimizer, model_dir, args.load_epoch)
if args.normal:
print('testing normal linemod ...')
image_db = LineModImageDB(args.linemod_cls,has_render_set=False,
has_fuse_set=False)
test_db = image_db.test_real_set+image_db.val_real_set
test_set = LineModDatasetRealAug(test_db, cfg.LINEMOD, vote_type, augment=False, use_motion=motion_model)
test_sampler = SequentialSampler(test_set)
test_batch_sampler = ImageSizeBatchSampler(test_sampler, train_cfg['test_batch_size'], False)
test_loader = DataLoader(test_set, batch_sampler=test_batch_sampler, num_workers=0)
prefix='test' if args.use_test_set else 'val'
val(net, test_loader, begin_epoch, prefix, use_motion=motion_model)
if args.occluded and args.linemod_cls in cfg.occ_linemod_cls_names:
print('testing occluded linemod ...')
occ_image_db = OcclusionLineModImageDB(args.linemod_cls)
occ_test_db = occ_image_db.test_real_set
occ_test_set = LineModDatasetRealAug(occ_test_db, cfg.OCCLUSION_LINEMOD, vote_type,
augment=False, use_motion=motion_model)
occ_test_sampler = SequentialSampler(occ_test_set)
occ_test_batch_sampler = ImageSizeBatchSampler(occ_test_sampler, train_cfg['test_batch_size'], False)
occ_test_loader = DataLoader(occ_test_set, batch_sampler=occ_test_batch_sampler, num_workers=0)
prefix='occ_test' if args.use_test_set else 'occ_val'
val(net, occ_test_loader, begin_epoch, prefix, use_motion=motion_model)
if args.truncated:
print('testing truncated linemod ...')
trun_image_db = TruncatedLineModImageDB(args.linemod_cls)
print(len(trun_image_db.set))
trun_image_set = LineModDatasetRealAug(trun_image_db.set, cfg.LINEMOD, vote_type, augment=False,
use_intrinsic=True, use_motion=motion_model)
trun_test_sampler = SequentialSampler(trun_image_set)
trun_test_batch_sampler = ImageSizeBatchSampler(trun_test_sampler, train_cfg['test_batch_size'], False)
trun_test_loader = DataLoader(trun_image_set, batch_sampler=trun_test_batch_sampler, num_workers=0)
prefix='trun_test'
val(net, trun_test_loader, begin_epoch, prefix, True, use_motion=motion_model)
else:
begin_epoch=0
if train_cfg['resume']:
begin_epoch=load_model(net.module.net, optimizer, model_dir)
image_db = LineModImageDB(args.linemod_cls,
has_fuse_set=train_cfg['use_fuse'],
has_render_set=True)
train_db=[]
train_db+=image_db.render_set
if train_cfg['use_real_train']:
train_db+=image_db.train_real_set
if train_cfg['use_fuse']:
train_db+=image_db.fuse_set
train_set = LineModDatasetRealAug(train_db, cfg.LINEMOD, vote_type, augment=True, cfg=train_cfg['aug_cfg'], use_motion=motion_model)
train_sampler = RandomSampler(train_set)
train_batch_sampler = ImageSizeBatchSampler(train_sampler, train_cfg['train_batch_size'], False, cfg=train_cfg['aug_cfg'])
train_loader = DataLoader(train_set, batch_sampler=train_batch_sampler, num_workers=12)
val_db=image_db.val_real_set
val_set = LineModDatasetRealAug(val_db, cfg.LINEMOD, vote_type, augment=False, cfg=train_cfg['aug_cfg'], use_motion=motion_model)
val_sampler = SequentialSampler(val_set)
val_batch_sampler = ImageSizeBatchSampler(val_sampler, train_cfg['test_batch_size'], False, cfg=train_cfg['aug_cfg'])
val_loader = DataLoader(val_set, batch_sampler=val_batch_sampler, num_workers=12)
if args.linemod_cls in cfg.occ_linemod_cls_names:
occ_image_db=OcclusionLineModImageDB(args.linemod_cls)
occ_val_db=occ_image_db.test_real_set[:len(occ_image_db.test_real_set)//2]
occ_val_set = LineModDatasetRealAug(occ_val_db, cfg.OCCLUSION_LINEMOD, vote_type, augment=False, cfg=train_cfg['aug_cfg'], use_motion=motion_model)
occ_val_sampler = SequentialSampler(occ_val_set)
occ_val_batch_sampler = ImageSizeBatchSampler(occ_val_sampler, train_cfg['test_batch_size'], False, cfg=train_cfg['aug_cfg'])
occ_val_loader = DataLoader(occ_val_set, batch_sampler=occ_val_batch_sampler, num_workers=12)
for epoch in range(begin_epoch, train_cfg['epoch_num']):
adjust_learning_rate(optimizer,epoch,train_cfg['lr_decay_rate'],train_cfg['lr_decay_epoch'])
train(net, optimizer, train_loader, epoch)
val(net, val_loader, epoch,use_motion=motion_model)
if args.linemod_cls in cfg.occ_linemod_cls_names:
val(net, occ_val_loader, epoch, 'occ_val',use_motion=motion_model)
save_model(net.module.net, optimizer, epoch, model_dir)
# def save_dataset(dataset,prefix=''):
# with open('assets/{}{}.txt'.format(prefix,args.linemod_cls),'w') as f:
# for data in dataset: f.write(data['rgb_pth']+'\n')
#
# def save_poses_dataset(prefix=''):
# print(np.asarray(poses_pr).shape)
# np.save('assets/{}{}_pr.npy'.format(prefix,args.linemod_cls),np.asarray(poses_pr))
# np.save('assets/{}{}_gt.npy'.format(prefix,args.linemod_cls),np.asarray(poses_gt))
if __name__ == "__main__":
train_net()
# save_poses_dataset('trun_')