-
Notifications
You must be signed in to change notification settings - Fork 6
/
generate_cloudseg.py
201 lines (171 loc) · 6.44 KB
/
generate_cloudseg.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
import argparse
import logging
import os
import subprocess
from pathlib import Path
import coloredlogs
from munch import Munch
from utilities.compressor import h264_compressor_cloudseg_segment
from utilities.results_utils import read_results
# qp_list = [32]
# gt_qp = 20
# qp_list = [20, 21, 22, 24, 26, 30, 34, 40]
# gt_qp = 50
# qp_list = [50, 51, 52, 53, 54, 56, 58, 60, 62]
# qp_list = [32, 42]
# quality_list = [
# "veryfast",
# "faster",
# "fast",
# "medium",
# "slow",
# "slower",
# "veryslow",
# ]
gt_qp = 30
attr = "mp4"
def main(args):
gt_qp = args.gt_qp
logger = logging.getLogger("mpeg_curve")
for video_name in args.inputs:
assert Path(video_name).is_dir()
video_name = Path(video_name)
# # generate ground truth
# output_names = []
# for quality in quality_list:
# input_name = f"{video_name}/%010d.png"
# output_name = f"{video_name}_qp_{gt_qp}_{quality}.hevc"
# output_names.append(output_name)
# print(f"Generate video for {output_name}")
# # encode_with_qp(input_name, output_name, qp, args)
# if args.force or not os.path.exists(output_name):
# # encode
# subprocess.run(
# [
# "ffmpeg",
# "-y",
# "-i",
# input_name,
# "-start_number",
# "0",
# "-qp",
# f"{gt_qp}",
# "-preset",
# f"{quality}",
# output_name,
# ]
# )
# # and inference
# subprocess.run(["python", "inference.py", "-i", output_name])
# subprocess.run(
# ["python", "merge_ground_truth.py", "-i"]
# + output_names
# + ["-o", f"{video_name}_qp_{gt_qp}_ground_truth.hevc"]
# )
# generate mpeg curve
for qp in args.qp_list:
new_args = Munch()
new_args.source = str(video_name)
new_args.output = str(video_name) + f"_cloudseg_qp_{qp}.mp4"
new_args.qp = qp
new_args.smooth_frames = args.smooth_frames
logger.info("Encode %s from %s", new_args.output, new_args.source)
# encode_with_qp(input_name, output_name, qp, args)
if args.force or not os.path.exists(new_args.output):
# if True:
h264_compressor_cloudseg_segment(new_args, logger)
subprocess.run(
[
"python",
"inference.py",
"-i",
new_args.output,
"--app",
args.app,
"--visualize_step_size",
"10000",
"--enable_cloudseg"
# "--confidence_threshold",
# "0.95",
]
)
subprocess.run(
[
"python",
"examine.py",
"-i",
new_args.output,
"-g",
f"{video_name}_qp_{gt_qp}.{attr}",
"--app",
args.app,
"--confidence_threshold",
f"{args.confidence_threshold}",
"--gt_confidence_threshold",
f"{args.gt_confidence_threshold}",
"--stats",
args.stats,
]
)
if __name__ == "__main__":
coloredlogs.install(
fmt="%(asctime)s [%(levelname)s] %(name)s:%(funcName)s[%(lineno)s] -- %(message)s",
level="INFO",
)
args = Munch()
# args.inputs = [
# "visdrone/videos/vis_%d" % i for i in [169, 170, 171, 172, 173, 209, 217]
# ]
# args.inputs = ["dashcam/dashcam_%d" % (i + 1) for i in [9]]
# args.inputs = [
# # "visdrone/videos/vis_171",
# # "visdrone/videos/vis_170",
# # "visdrone/videos/vis_173",
# # "visdrone/videos/vis_169",
# # "visdrone/videos/vis_172",
# ]
# args.inputs = ["DAVIS/videos/DAVIS_1"]
# args.inputs = ["visdrone/videos/vis_%d" % i for i in range(169, 174)] + [
# "dashcam/dashcam_%d" % i for i in range(1, 11)
# ]
# args.inputs = ["dashcam/dashcam_%d" % i for i in range(5, 11)]
# args.inputs = ["adapt/drive_%d" % i for i in range(60)]
# args.inputs = ["visdrone/videos/vis_%d" % i for i in [169]]
# args.inputs = ["large_object/large_%d" % i for i in range(1, 5)]
# args.inputs = ["dashcam/dashcam_%d" % i for i in range(4, 11)]
# args.inputs = ["dashcam/dashcam_%d" % i for i in range(1, 8)]
# args.inputs = ["dashcam/dashcam_%d" % i for i in [2, 5, 6, 8]]
# args.inputs = ["visdrone/videos/vis_171"]
# args.gt_qp = 20
# args.qp_list = [20, 21, 22, 24, 26, 30, 34, 40]
args.gt_qp = 30
# args.qp_list = [30, 31, 32, 34, 36, 40, 44, 50]
# args.qp_list = args.qp_list + [33, 35, 37, 38, 39]
args.qp_list = [26]
# args.qp_list = [30]
# args.qp_list = [20, 27, 28, 30, 32, 34, 35, 36, 38, 40, 46]
# args.inputs = [
# "visdrone/videos/vis_%d" % i for i in [169, 170, 171, 172, 173]
# ]
# args.inputs = ["yoda/yoda_%d" % i for i in range(7, 8)]
# args.inputs = ["dashcam/dashcamcropped_%d" % i for i in range(1, 11)]
args.inputs = ["videos/driving_%d" % i for i in range(5)] + [
"videos/dashcamcropped_%d" % i for i in range(1, 11)
]
# args.inputs = ["videos/driving_%d" % i for i in range(5)]
args.inputs = [args.inputs[i] for i in range(len(args.inputs))]
args.force = False
# args.app = "COCO-Detection/faster_rcnn_R_101_FPN_3x.yaml"
# args.app = "Yolo5s"
args.app = "EfficientDet"
# assert attr == "webm"
# args.stats = f"frozen_stats_MLSys/stats_QP30_thresh7_segmented_FPN"
# args.stats = "frozen_stats_MLSys/stats_QP30_thresh3_segment_Yolo"
args.stats = "frozen_stats_MLSys/stats_QP30_thresh4_segment_EfficientDet"
# args.stats = "frozen_stats_MLSys/stats_QP30_thresh3_dashcamcropped_Yolo"
# args.stats = "frozen_stats_MLSys/stats_QP30_thresh4_dashcamcropped_EfficientDet"
args.confidence_threshold = 0.4
args.gt_confidence_threshold = 0.4
args.smooth_frames = 10
# args = parser.parse_args()
main(args)