forked from ZQPei/deep_sort_pytorch
-
Notifications
You must be signed in to change notification settings - Fork 0
/
yolov3_deepsort.py
204 lines (166 loc) · 7.9 KB
/
yolov3_deepsort.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
import os
import cv2
import time
import argparse
import torch
import warnings
import pickle
import re
import numpy as np
from detector import build_detector
from deep_sort import build_tracker
from utils.draw import draw_boxes
from utils.parser import get_config
from utils.log import get_logger
from utils.io import write_results
class VideoTracker(object):
def __init__(self, cfg, args, video_path):
self.cfg = cfg
self.args = args
self.video_path = video_path
self.logger = get_logger("root")
self.logger.info(f"Video rate is {self.video_path}")
use_cuda = args.use_cuda and torch.cuda.is_available()
if not use_cuda:
warnings.warn("Running in cpu mode which maybe very slow!", UserWarning)
if args.display:
cv2.namedWindow("test", cv2.WINDOW_NORMAL)
cv2.resizeWindow("test", args.display_width, args.display_height)
if args.cam != -1:
print("Using webcam " + str(args.cam))
self.vdo = cv2.VideoCapture(args.cam)
else:
self.vdo = cv2.VideoCapture()
# Uniformly sample frames to save resources
# Copy the previous result when a frame is skipped
self.logger.info(f"Sample rate is {args.sample_rate}")
self.skip_frame = int(1 / args.sample_rate)
self.logger.info(f"Detection model is set to {args.detection_model}")
self.detector = build_detector(args.detection_model, cfg, use_cuda=use_cuda)
self.deepsort = build_tracker(cfg, use_cuda=use_cuda)
self.class_names = self.detector.class_names
self.temp_tesult = []
def __enter__(self):
if self.args.cam != -1:
ret, frame = self.vdo.read()
assert ret, "Error: Camera error"
self.im_width = frame.shape[0]
self.im_height = frame.shape[1]
else:
assert os.path.isfile(self.video_path), "Path error"
self.vdo.open(self.video_path)
self.im_width = int(self.vdo.get(cv2.CAP_PROP_FRAME_WIDTH))
self.im_height = int(self.vdo.get(cv2.CAP_PROP_FRAME_HEIGHT))
assert self.vdo.isOpened()
if self.args.save_path:
os.makedirs(self.args.save_path, exist_ok=True)
# path of saved video and results
self.save_video_path = os.path.join(self.args.save_path, self.args.save_file + ".avi")
self.save_results_path = os.path.join(self.args.save_path, self.args.save_file + ".txt")
# create video writer
fourcc = cv2.VideoWriter_fourcc(*'MJPG')
if args.force_resolution:
self.writer = cv2.VideoWriter(self.save_video_path, fourcc, 20, (1920, 1080))
else:
self.writer = cv2.VideoWriter(self.save_video_path, fourcc, 20, (self.im_width, self.im_height))
# logging
self.logger.info(f"Saving video to {self.save_video_path}")
self.logger.info(f"Saving result to {self.save_results_path}")
return self
def __exit__(self, exc_type, exc_value, exc_traceback):
if exc_type:
print(exc_type, exc_value, exc_traceback)
def get_next_detection(self, image, idx_frame):
# Sample frames to save resources
# self.logger.info(f"skip frame is {self.skip_frame}")
if idx_frame % self.skip_frame == 0:
self.logger.info(f"Running detection for frame {idx_frame}")
self.temp_tesult = self.detector(image)
return self.temp_tesult
def run(self):
results = []
fps = []
idx_frame = 0
detection_dict = {}
while self.vdo.grab():
# if idx_frame % self.args.frame_interval:
# continue
start = time.time()
_, ori_im = self.vdo.retrieve()
im = cv2.cvtColor(ori_im, cv2.COLOR_BGR2RGB)
height, width = im.shape[:2]
# do detection
bbox_xywh, cls_conf, cls_ids = self.get_next_detection(im, idx_frame)
# print(bbox_xywh, cls_conf, cls_ids)
# # if idx_frame == 3:
# # break
if args.save_detection:
detection_dict[idx_frame] = [bbox_xywh, cls_conf, cls_ids]
idx_frame += 1
# select person class
mask = cls_ids == 0
bbox_xywh = bbox_xywh[mask]
# bbox dilation just in case bbox too small, delete this line if using a better pedestrian detector
bbox_xywh[:, 3:] *= 1.2
cls_conf = cls_conf[mask]
# do tracking
outputs = self.deepsort.update(bbox_xywh, cls_conf, im)
# draw boxes for visualization
if len(outputs) > 0:
bbox_tlwh = []
bbox_xyxy = outputs[:, :4]
# x_scale = int(np.round(1920/width))
# y_scale = int(np.round(1080/height))
# x_scale = 1920/width
# y_scale = 1080/height
# bbox_xyxy *= np.array([[x_scale, y_scale, x_scale, y_scale]], dtype=np.int32)
identities = outputs[:, -1]
ori_im = draw_boxes(ori_im, bbox_xyxy, identities, force_resolution=args.force_resolution)
for bb_xyxy in bbox_xyxy:
bbox_tlwh.append(self.deepsort._xyxy_to_tlwh(bb_xyxy))
results.append((idx_frame - 1, bbox_tlwh, identities))
end = time.time()
if self.args.display:
cv2.imshow("test", ori_im)
cv2.waitKey(1)
if self.args.save_path:
self.writer.write(ori_im)
# save results
write_results(self.save_results_path, results, 'mot')
# logging
fps.append(1 / (end - start))
self.logger.info("time: {:.03f}s, fps: {:.03f}, detection numbers: {}, tracking numbers: {}" \
.format(end - start, 1 / (end - start), bbox_xywh.shape[0], len(outputs)))
self.logger.info("Average fps is {:.03f}".format(sum(fps) / len(fps)))
if args.save_detection:
res = re.findall(r'\d+', self.video_path)[0]
save_path = str(args.detection_model) + "_" + str(args.sample_rate).replace(".", "") + "_" + res + ".pkl"
self.logger.info(f"Saving detection results to {save_path}")
with open(os.path.join("output", save_path), "wb") as f:
pickle.dump(detection_dict, f)
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument("VIDEO_PATH", type=str)
parser.add_argument("--detection_model", type=str, default="yolov3")
parser.add_argument("--sample_rate", type=float, default=1.0)
parser.add_argument("--config_detection", type=str, default="./configs/yolov3.yaml")
parser.add_argument("--config_deepsort", type=str, default="./configs/deep_sort.yaml")
# parser.add_argument("--ignore_display", dest="display", action="store_false", default=True)
parser.add_argument("--display", action="store_true")
# parser.add_argument("--frame_interval", type=int, default=1)
parser.add_argument("--display_width", type=int, default=800)
parser.add_argument("--display_height", type=int, default=600)
parser.add_argument("--save_path", type=str, default="./output/")
parser.add_argument("--save_file", type=str, default="results")
parser.add_argument("--save_detection", type=bool, default=False)
parser.add_argument("--force_resolution", type=bool, default=False)
parser.add_argument("--cpu", dest="use_cuda", action="store_false", default=True)
parser.add_argument("--camera", action="store", dest="cam", type=int, default="-1")
return parser.parse_args()
if __name__ == "__main__":
args = parse_args()
cfg = get_config()
cfg.merge_from_file(args.config_detection)
cfg.merge_from_file(args.config_deepsort)
with VideoTracker(cfg, args, video_path=args.VIDEO_PATH) as vdo_trk:
vdo_trk.run()