-
Notifications
You must be signed in to change notification settings - Fork 1
/
objection_detection_app.py
223 lines (175 loc) · 7.64 KB
/
objection_detection_app.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
import cv2 as cv
import argparse
import numpy as np
import tensorflow as tf
import os
#Python Libraries
from queue import Queue #Thread safe
from threading import Thread
from copy import deepcopy
#Local Files
from utils.notifier import notifier
from utils.tracking import ObjectTracker
from utils.detect_object import detect_objects
from utils.videoContour import video2Contour
"""
Usage example:
Webcam: python object_detection_app.py
Video: python object_detection_app.py --video run.mp4
Image: python object_detection_app.py --image bird.jpg
Models:
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md
Optimizing Only for python 2:
https://www.pyimagesearch.com/2015/12/21/increasing-webcam-fps-with-python-and-opencv/
https://www.pyimagesearch.com/2017/02/06/faster-video-file-fps-with-cv2-videocapture-and-opencv/
"""
"""
This thread carries out object detection and loads up the frozen tensorflow model for
classification.
Takes the frames from input_queue calculates classification data and pushes
data and original frame to output_queue
Input:
-input_q: Input queue, contains frames
-output_q: Output queue, original frame + data will be pushed to this queue
Return: None
"""
def thread_detect_objects(input_q, output_q):
# load the frozen tensorflow model into memory
detection_graph = tf.Graph()
with detection_graph.as_default():
od_graph_def = tf.GraphDef()
with tf.gfile.GFile(PATH_TO_MODEL, 'rb') as fid:
serialized_graph = fid.read()
od_graph_def.ParseFromString(serialized_graph)
tf.import_graph_def(od_graph_def, name='')
sess = tf.Session(graph=detection_graph)
while not kill_threads:
if not input_q.empty(): #better to block than poll empty, as not running on various cores
frame = input_q.get()
data = detect_objects(frame, sess, detection_graph)
output_q.put(data)
output_q.put(frame)
sess.close()
"""
Tried to optimize with threads but ended up working slower, is it able to run with various cores?
https://stackoverflow.com/questions/7542957/is-python-capable-of-running-on-multiple-cores
Base on various answer it's unable to run on various cores
If future python version work correctly on multiple cores this will result useful, use one thread to process
and one to output video file.
"""
def thread_process_image(input_q, process_q):
return
def thread_output_image(process_q, output_q):
return
"""
Carries out the parsing of the command line, this will detect if the user wants to process webcam,
images or videos. Returns video capture source to be used with cv.VideoCapture() and the name of the output file
Input: None
Return:
-video_capture_source: 0 for webcam, and for image or video files just the file name
-output_file: Returns output file name, "...out_py.avi" for video, "...out_py.jpg" and "NULL" for webcam a
there shouldn't be an output file
-save_vid: Returns true if save video option selected, will store video
-show_vid: Returns true if show video option selected, will show video in window while processing
"""
def parse_cmd_line():
video_capture_source = 0 #Webcam number
output_file = "NULL"
parser = argparse.ArgumentParser(description ='Real Time Object Detection using OPENCV + TF')
group_source = parser.add_mutually_exclusive_group()
group_source.add_argument('-img', '--image', help='Path to image file.')
group_source.add_argument('-vid', '--video', help='Path to video file.')
group_source.add_argument('-ip', '--ipcam', help='Ipcam Ip number.')
parser.add_argument('-sv', '--save', help='Saves videos to memory.', action="store_true")
parser.add_argument('-sh', '--show', help='Shows videos in windows while processing.', action="store_true")
#parser.add_argument('-o', '--output', help='Output name.')
args = parser.parse_args()
save_vid = args.save
show_vid = args.show
if not save_vid and not show_vid: #Default show video
show_vid = True
print("No option choosen, will default show video. WILL NOT BE SAVED TO MEM -h for more options")
if(args.video):
# Open the video file
if not os.path.isfile(args.video):
print("Input video file ", args.video, " doesn't exist")
sys.exit(1)
output_file = args.video[:-4]+'_video_out_py.avi' #Removes extension and adds _video_out_py.ave
video_capture_source = args.video
elif(args.ipcam):
output_file = args.ipcam +'_ipcam_out_py.avi'
video_capture_source = args.ipcam
elif(args.image):
#Open image file
if not os.path.isfile(args.image):
print("Input image file ", args.image, " doesn't exist")
sys.exit(1)
video_capture_source = args.image
output_file = args.image[:-4]+'_image_out_py.jpg'
return video_capture_source, output_file, save_vid, show_vid
# DEFINES
CWD_PATH = os.getcwd()
MODEL_NAME = 'ssdlite_mobilenet_v2_coco_2018_05_09'
PATH_TO_MODEL = os.path.join(CWD_PATH, 'detection', MODEL_NAME, 'frozen_inference_graph.pb')
WINDOW_NAME = "People Detector"
#Global Variables
kill_threads = False #will be used to kill threads
if __name__ == '__main__':
video_capture_source, output_file, save_vid, show_vid = parse_cmd_line()
split_view = False #When S key is pressed will split the view in half of processed non processed
do_exit = False #Will be used to kill the loop
pending_frames = 0 #Will carry out a count of pending
is_image = (output_file[-4:] =='.jpg') #Incase user input -video and -image video will have priority
#Designed for multithreading but python 3 doesn't support various cores running D:
input_q = Queue(1)
output_q = Queue()
object_tracker = ObjectTracker()
video_contour = video2Contour()
website_html = notifier(15)
Thread(target=thread_detect_objects, args=(input_q, output_q)).start()
#Viceo_capture_source integer corresponds internal cam, while string corresponds to file path or ipcam
vid_capture = cv.VideoCapture(video_capture_source)
width = round(vid_capture.get(cv.CAP_PROP_FRAME_WIDTH))
height = round(vid_capture.get(cv.CAP_PROP_FRAME_HEIGHT))
codec = cv.VideoWriter_fourcc(*'MJPG')#http://www.fourcc.org/codecs.php
#Vid writer is necessary for saving a video file
if save_vid:
vid_writer = cv.VideoWriter(output_file, codec , 30, (width,height))
while not do_exit : #If q key is pressed exit window
get_key = cv.waitKey(1) & 0xFF
if get_key == ord('q') or get_key == 27: # q or 27==Esc key to stop
do_exit = True
elif get_key == ord('s'):
split_view ^= True
has_frame, frame = vid_capture.read() #has_frame returns false when reaching end of file
if has_frame:# If not end frame put into the input queue
input_q.put(frame)
pending_frames += 1
elif pending_frames == 0: #No more pending frames
break
if not output_q.qsize() >= 2: #Check if empty to avoid blocking
data = output_q.get()
new_frame = output_q.get()
context = {'frame': new_frame, 'class_names': data['class_names'], 'rec_points': data['rect_points'],
'class_colors': data['class_colors'], 'width': width, 'height': height}
website_html(deepcopy(context)) #Avoid the frame from being modified
new_frame = object_tracker(context)
pending_frames -= 1
if split_view:
processed_frame = video_contour.apply(deepcopy(new_frame))
#Leave half with original frame and half with processed_frame
new_frame[:, :int(context['width']//2)] = cv.cvtColor(processed_frame[:, :int(context['width']//2)] ,cv.COLOR_GRAY2BGR)
if save_vid:
vid_writer.write(new_frame.astype(np.uint8))
if show_vid:
cv.imshow(WINDOW_NAME, new_frame)
if is_image:
cv.imwrite(output_file, new_frame.astype(np.uint8));
kill_threads = True
vid_capture.release()
if save_vid or is_image:
print("Done processing!")
print("Output file stored as: ", output_file)
if save_vid: #0 = Webcam
vid_writer.release()
cv.destroyAllWindows()