-
Notifications
You must be signed in to change notification settings - Fork 3
/
Flownet2ControllerFineTune.py
266 lines (214 loc) · 10.2 KB
/
Flownet2ControllerFineTune.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
import torch
from torch.autograd import Variable
import argparse
import numpy as np
from os.path import *
import cv2
import flownet2.models as models
import flownet2.losses as losses
import flownet2.datasets as datasets
from flownet2.utils import tools
from finetuned.loss import unsupervised_loss
from finetuned.interpolations import warp_unidirectional_flow
parser = argparse.ArgumentParser()
parser.add_argument("--start_epoch", type=int, default=1)
parser.add_argument("--total_epochs", type=int, default=10000)
parser.add_argument("--batch_size", "-b", type=int, default=8, help="Batch size")
parser.add_argument(
"--train_n_batches",
type=int,
default=-1,
help="Number of min-batches per epoch. If < 0, it will be determined by training_dataloader",
)
parser.add_argument(
"--crop_size",
type=int,
nargs="+",
default=[256, 256],
help="Spatial dimension to crop training samples for training",
)
parser.add_argument("--gradient_clip", type=float, default=None)
parser.add_argument("--schedule_lr_frequency", type=int, default=0, help="in number of iterations (0 for no schedule)")
parser.add_argument("--schedule_lr_fraction", type=float, default=10)
parser.add_argument("--rgb_max", type=float, default=255.0)
parser.add_argument("--number_workers", "-nw", "--num_workers", type=int, default=8)
parser.add_argument("--number_gpus", "-ng", type=int, default=-1, help="number of GPUs to use")
parser.add_argument("--no_cuda", action="store_true")
parser.add_argument("--seed", type=int, default=1)
parser.add_argument("--name", default="run", type=str, help="a name to append to the save directory")
parser.add_argument("--save", "-s", default="./work", type=str, help="directory for saving")
parser.add_argument("--validation_frequency", type=int, default=5, help="validate every n epochs")
parser.add_argument("--validation_n_batches", type=int, default=-1)
parser.add_argument(
"--render_validation",
action="store_true",
help="run inference (save flows to file) and every validation_frequency epoch",
)
parser.add_argument("--inference", action="store_true")
parser.add_argument(
"--inference_size",
type=int,
nargs="+",
default=[-1, -1],
help="spatial size divisible by 64. default (-1,-1) - largest possible valid size would be used",
)
parser.add_argument("--inference_batch_size", type=int, default=1)
parser.add_argument("--inference_n_batches", type=int, default=-1)
parser.add_argument("--save_flow", action="store_true", help="save predicted flows to file")
parser.add_argument("--resume", default="", type=str, metavar="PATH", help="path to latest checkpoint (default: none)")
parser.add_argument("--log_frequency", "--summ_iter", type=int, default=1, help="Log every n batches")
parser.add_argument("--skip_training", action="store_true")
parser.add_argument("--skip_validation", action="store_true")
parser.add_argument("--fp16", action="store_true", help="Run model in pseudo-fp16 mode (fp16 storage fp32 math).")
parser.add_argument(
"--fp16_scale",
type=float,
default=1024.0,
help="Loss scaling, positive power of 2 values can improve fp16 convergence.",
)
tools.add_arguments_for_module(parser, models, argument_for_class="model", default="FlowNet2")
tools.add_arguments_for_module(parser, losses, argument_for_class="loss", default="L1Loss")
tools.add_arguments_for_module(
parser, torch.optim, argument_for_class="optimizer", default="Adam", skip_params=["params"]
)
tools.add_arguments_for_module(
parser,
datasets,
argument_for_class="training_dataset",
default="MpiSintelFinal",
skip_params=["is_cropped"],
parameter_defaults={"root": "./MPI-Sintel/flow/training"},
)
tools.add_arguments_for_module(
parser,
datasets,
argument_for_class="validation_dataset",
default="MpiSintelClean",
skip_params=["is_cropped"],
parameter_defaults={"root": "./MPI-Sintel/flow/training", "replicates": 1},
)
tools.add_arguments_for_module(
parser,
datasets,
argument_for_class="inference_dataset",
default="MpiSintelClean",
skip_params=["is_cropped"],
parameter_defaults={"root": "./MPI-Sintel/flow/training", "replicates": 1},
)
args = parser.parse_args()
if args.number_gpus < 0:
args.number_gpus = torch.cuda.device_count()
parser.add_argument("--IGNORE", action="store_true")
defaults = vars(parser.parse_args(["--IGNORE"]))
args.model_class = tools.module_to_dict(models)[args.model]
args.optimizer_class = tools.module_to_dict(torch.optim)[args.optimizer]
args.loss_class = tools.module_to_dict(losses)[args.loss]
args.training_dataset_class = tools.module_to_dict(datasets)[args.training_dataset]
args.validation_dataset_class = tools.module_to_dict(datasets)[args.validation_dataset]
args.inference_dataset_class = tools.module_to_dict(datasets)[args.inference_dataset]
args.cuda = not args.no_cuda and torch.cuda.is_available()
args.log_file = join(args.save, "args.txt")
args.grads = {}
if args.inference:
args.skip_validation = True
args.skip_training = True
args.total_epochs = 1
args.inference_dir = "{}/inference".format(args.save)
args.effective_batch_size = args.batch_size * args.number_gpus
args.effective_inference_batch_size = args.inference_batch_size * args.number_gpus
args.effective_number_workers = args.number_workers * args.number_gpus
gpuargs = {"num_workers": args.effective_number_workers, "pin_memory": True, "drop_last": True} if args.cuda else {}
inf_gpuargs = gpuargs.copy()
inf_gpuargs["num_workers"] = args.number_workers
class FlowControllerFineTune:
def __init__(self, model_path="./flownet2/pretrained_models/FlowNet2_checkpoint.pth.tar"):
self.model = models.FlowNet2(args)
checkpoint = torch.load(model_path)
self.model.load_state_dict(checkpoint["state_dict"])
self.model.eval()
if torch.cuda.is_available():
self.model.cuda()
self.is_cropped = False
@staticmethod
def convert_flow_to_image(flow):
image_shape = flow.shape[0:2] + (3,)
hsv = np.zeros(shape=image_shape, dtype=np.uint8)
hsv[..., 1] = 255
mag, ang = cv2.cartToPolar(flow[..., 0], flow[..., 1])
hsv[..., 0] = ang * 180 / np.pi / 2
normalized_mag = np.asarray(np.clip(mag * 40, 0, 255), dtype=np.uint8)
hsv[..., 2] = normalized_mag
rgb = cv2.cvtColor(hsv, cv2.COLOR_HSV2RGB)
rgb = np.asarray(rgb, np.uint8)
return rgb
def convert_video_to_flow(self, video_path, output_path="out", downsample_res=None, raw_save=False):
video = cv2.VideoCapture(video_path)
ret, prev_frame = video.read()
if downsample_res is not None:
prev_frame = cv2.resize(prev_frame, downsample_res)
view_shape = list(prev_frame.shape[0:2])
if not raw_save:
view_shape[0] *= 2
out_video = cv2.VideoWriter(
output_path + ".avi", cv2.VideoWriter_fourcc("M", "J", "P", "G"), 24, tuple(view_shape)
)
while video.isOpened():
ret, frame = video.read()
if ret == True:
if downsample_res is not None:
frame = cv2.resize(frame, downsample_res)
opt_flow = self.predict(frame, prev_frame)
opt_flow_image = self.convert_flow_to_image(opt_flow)
prev_frame = frame
joint_image = np.append(frame, opt_flow_image, axis=1)
cv2.imshow("FlowNet2", joint_image)
if raw_save:
out_video.write(opt_flow_image)
else:
out_video.write(joint_image)
if cv2.waitKey(1) & 0xFF == ord("q"):
break
else:
break
video.release()
out_video.release()
cv2.destroyAllWindows()
@staticmethod
def preprocess_frames(frame1, frame2):
assert frame1.shape == frame2.shape, "Shapes of both frames must be same"
# Downscale image resolution to closest factor for 64, if smaller than 64 than upscale to 64
# This part basically calculates which resolution it should scale the image to
process_resolution = tuple([max(64 * (frame1.shape[i] // 64), 64) for i in range(2)])
images = [cv2.resize(frame1, process_resolution), cv2.resize(frame2, process_resolution)]
images = np.expand_dims(np.array(images).transpose(3, 0, 1, 2), axis=0)
images = torch.from_numpy(images.astype(np.float32))
return [images], [torch.zeros(images.size()[0:1] + (2,) + images.size()[-2:])]
def predict(self, image1, image2):
(data, target) = self.preprocess_frames(image1, image2)
if args.cuda:
data, target = [d.cuda() for d in data], [t.cuda() for t in target]
data, target = [Variable(d) for d in data], [Variable(t) for t in target]
with torch.no_grad():
output = self.model(data[0])
flow = cv2.resize(output.data.cpu().numpy()[0].transpose(1, 2, 0), (image1.shape[1], image1.shape[0]))
return flow
def fine_tune(self, image1, image2, num_iter=1, lr=1e-4):
(data, target) = self.preprocess_frames(image1, image2)
if args.cuda:
data, target = [d.cuda() for d in data], [t.cuda() for t in target]
data, target = [Variable(d) for d in data], [Variable(t) for t in target]
optimizer = torch.optim.Adam(self.model.parameters(), lr=lr)
self.model.train()
for _ in range(num_iter):
output = self.model(data[0])
optical_flow = cv2.resize(
output.data.cpu().numpy()[0].transpose(1, 2, 0), (image1.shape[1], image1.shape[0])
)
flow = [optical_flow[:, :, 0], optical_flow[:, :, 1]]
warped_image = warp_unidirectional_flow(image2, flow)
loss = unsupervised_loss(flow, warped_image, image1)
if optimizer is not None:
optimizer.zero_grad()
loss.backward()
optimizer.step()
self.model.eval()