-
Notifications
You must be signed in to change notification settings - Fork 2
/
evaluate.py
240 lines (181 loc) · 9.23 KB
/
evaluate.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
import sys
sys.path.append('core')
import argparse
import os
import os.path as osp
import numpy as np
import torch
import datasets
from utils import frame_utils
from utils.utils import InputPadder, forward_interpolate
from CCMR import CCMR
from config.config_loader import cpy_eval_args_to_config
from custom_logger import init_logger
import logging
from tqdm import tqdm
@torch.no_grad()
def create_kitti_submission(model, iters, output_path='kitti_submission'):
""" Create submission for the Sintel leaderboard """
model.eval()
test_dataset = datasets.KITTI(split='testing', aug_params=None)
coarsest_scale = 16
if not os.path.exists(output_path):
os.makedirs(output_path)
for test_id in tqdm(range(len(test_dataset))):
image1, image2, (frame_id, ) = test_dataset[test_id]
padder = InputPadder(image1.shape, mode='kitti', coarsest_scale=coarsest_scale)
image1, image2 = padder.pad(image1[None].cuda(), image2[None].cuda())
_, flow_pr = model(image1=image1, image2=image2, iters=iters, test_mode=True)
flow = padder.unpad(flow_pr[0]).permute(1, 2, 0).cpu().numpy()
output_filename = os.path.join(output_path, frame_id)
frame_utils.writeFlowKITTI(output_filename, flow)
@torch.no_grad()
def create_sintel_submission(model, iters, output_path='sintel_submission', split= "test"):
""" Create submission for the Sintel leaderboard """
model.eval()
coarsest_scale = 16
for dstype in ['final', 'clean']:
test_dataset = datasets.MpiSintel(split=split, aug_params=None, dstype=dstype, show_extra_info=True)
flow_prev, sequence_prev = None, None
for test_id in tqdm(range(len(test_dataset))):
if split == "test":
image1, image2, (sequence, frame) = test_dataset[test_id]
elif split == "training":
image1, image2,_,_, (sequence, frame) = test_dataset[test_id]
if sequence != sequence_prev:
flow_prev = None
padder = InputPadder(image1.shape, coarsest_scale=coarsest_scale)
image1, image2 = padder.pad(image1[None].cuda(), image2[None].cuda())
_, flow_pr = model(image1=image1, image2=image2, iters=iters, flow_init=flow_prev, test_mode=True)#cold warm
flow = padder.unpad(flow_pr[0]).permute(1, 2, 0).cpu().numpy()
init_flow_for_next_low, _ = model(image1=image1, image2=image2, iters=iters, flow_init=None, test_mode=True)#cold warm
flow_prev = forward_interpolate(init_flow_for_next_low[0])[None].cuda()#cold warm
output_dir = os.path.join(output_path, dstype, sequence)
output_file = os.path.join(output_dir, 'frame%04d.flo' % (frame+1))
if not os.path.exists(output_dir):
os.makedirs(output_dir)
frame_utils.writeFlow(output_file, flow)
sequence_prev = sequence
@torch.no_grad()
def validate_sintel(model, iters, model_path):
""" Evaluate trained model on Sintel(train) clean + final passes on train split """
logger = logging.getLogger('eval.sintel')
logger_for_eval_all = logging.getLogger('eval_all.sintel')
model.eval()
results = {}
coarsest_scale = 16
epe_sequence={}
for dstype in ['clean', 'final']:
val_dataset = datasets.MpiSintel(split='training', dstype=dstype, show_extra_info = True)
epe_list = []
efel_list = []
flow_prev, sequence_prev = None, None
sequences = []
for val_id in tqdm(range(len(val_dataset))):
image1, image2, flow_gt, _, (sequence, frame) = val_dataset[val_id]
frame += 1
if sequence not in sequences:
sequences.append(sequence)
if f"{sequence}_epe" not in epe_sequence:
epe_sequence[f"{sequence}_epe"] = 0
epe_sequence[f"{sequence}_count"] = 0
padder = InputPadder(image1.shape, coarsest_scale=coarsest_scale)
image1, image2 = padder.pad(image1[None].cuda(), image2[None].cuda())
if sequence != sequence_prev:
flow_prev = None
flow_low, flow_pr = model(image1, image2, iters=iters, flow_init=flow_prev, test_mode=True)
flow = padder.unpad(flow_pr[0]).cpu()
flow_prev = forward_interpolate(flow_low[0])[None].cuda()
epe = torch.sum((flow - flow_gt)**2, dim=0).sqrt()
epe_list.append(epe.view(-1).numpy())
#*****
mag = torch.sum(flow_gt**2, dim=0).sqrt()
epe = epe.view(-1)
mag = mag.view(-1)
efel = ((epe > 3.0) & ((epe/mag) > 0.05)).float()
efel_list.append(efel.cpu().numpy())
#******
#logger.info("%s %s %d EPE: %f, " % (dstype, sequence, frame, epe.mean()))
sequence_prev = sequence
epe_sequence[f"{sequence}_epe"] += epe.mean()
epe_sequence[f"{sequence}_count"] += 1
for seq in sequences:
epe_per_seq = epe_sequence[f"{seq}_epe"]/epe_sequence[f"{seq}_count"]
logger.info("Sintel validation, iters: %d, type: %s seq:(%s) EPE: %f" % (sum([8, 10, 10, 10]), dstype, seq, epe_per_seq))
efel_list = np.concatenate(efel_list)
FL = 100 * np.mean(efel_list)
epe_all = np.concatenate(epe_list)
epe = np.mean(epe_all)
px1 = np.mean(epe_all<1)
px3 = np.mean(epe_all<3)
px5 = np.mean(epe_all<5)
logger.info("Sintel validation iters:%d (%s) EPE: %f, 1px: %f, 3px: %f, 5px: %f, FL-Error: %f:" % (sum([8, 10, 10, 10]), dstype, epe, px1, px3, px5, FL))
logger_for_eval_all.info("Sintel iters:%d validation of the model: %s , : (%s) EPE: %f , FL-Error: %f:" % (sum([8, 10, 10, 10]), model_path, dstype, epe, FL))
results[dstype] = np.mean(epe_list)
return results
@torch.no_grad()
def validate_kitti(model, iters, padding="kitti"):
""" Peform validation using the KITTI-2015 (train) split """
logger = logging.getLogger("eval.kitti")
model.eval()
val_dataset = datasets.KITTI(split='training')
out_list, epe_list = [], []
coarsest_scale = 16
for data_id in tqdm(range(len(val_dataset))): # uses len() and updates correctly
image1, image2, flow_gt, valid_gt = val_dataset[data_id]
image1 = image1[None].cuda()
image2 = image2[None].cuda()
padder = InputPadder(image1.shape, mode=padding, coarsest_scale=coarsest_scale)
image1, image2 = padder.pad(image1, image2)
flow_low, flow_pr = model(image1, image2, iters=iters, test_mode=True)
flow = padder.unpad(flow_pr[0]).cpu()
epe = torch.sum((flow - flow_gt)**2, dim=0).sqrt()
mag = torch.sum(flow_gt**2, dim=0).sqrt()
epe = epe.view(-1)
mag = mag.view(-1)
val = valid_gt.view(-1) >= 0.5
out = ((epe > 3.0) & ((epe/mag) > 0.05)).float()
epe_list.append(epe[val].mean().item())
out_list.append(out[val].cpu().numpy())
epe_list = np.array(epe_list)
out_list = np.concatenate(out_list)
epe = np.mean(epe_list)
f1 = 100 * np.mean(out_list)
logger.warning("Kitti validation:%f, %f" % (epe, f1))
return {'kitti-epe': epe, 'kitti-f1': f1}
def str2bool(v):
return v.lower() in ("yes", "true", "t", "1", "True")
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--model', help="restore checkpoint")
parser.add_argument('--model_type', choices=['CCMR', 'CCMR+'], default="CCMR+", help="Choose \"CCMR\" or \"CCMR+\" for the 3-scale and 4-scale version, respectively.")
parser.add_argument('--dataset', help="dataset for evaluation")
parser.add_argument('--num_scales', type=int, help="3 for CCMR and 4 for CCMR+", default=4)
parser.add_argument('--mixed_precision', help='use mixed precision', type = str2bool, default = True)
args = parser.parse_args()
config = cpy_eval_args_to_config(args)
if config["model_type"] == "CCMR":
sintel_iters = [8, 10, 15]
kitti_iters = [6, 8, 30]
config["num_scales"] = 3
elif config["model_type"] == "CCMR+":
sintel_iters = [8, 10, 10, 10]
kitti_iters = [35, 35, 5, 15]
config["num_scales"] = 4
model = torch.nn.DataParallel(CCMR(config))
model.load_state_dict(torch.load(config["model"]))
path = os.path.join(os.path.dirname((config["model"])), "eval.log")
logger = init_logger("eval", path, stream_level= logging.WARNING)
logger = init_logger("eval_all", "./eval_all.log")
model.cuda()
model.eval()
with torch.no_grad():
if config["dataset"] == 'sintel':
validate_sintel(model=model.module, iters= sintel_iters, model_path=config["model"])
elif config["dataset"] == 'sintel_test':
out_path = osp.join(os.path.dirname(config["model"]), 'sintel_test')
create_sintel_submission(model, iters=sintel_iters, output_path=out_path)
elif config["dataset"] == 'kitti':
validate_kitti(model.module, iters=kitti_iters)
elif config["dataset"] == "kitti_test":
create_kitti_submission(model, iter=kitti_iters, output_path=osp.join(os.path.dirname(config["model"]), 'kitti_test', 'flow'))