-
Notifications
You must be signed in to change notification settings - Fork 3
/
video_classifier.py
170 lines (147 loc) · 6.3 KB
/
video_classifier.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
import os
import time
import cv2
#import couchdb
import numpy as np
from threading import Thread
from skimage.measure import compare_ssim as ssim
from db_handler import *
import subprocess
import shlex
import json
import datetime
import resource
class video_classifier():
def __init__(self, files, tags = None):
self.media = files
self.results = []
self.times = []
self.results = []
self.tags = tags
self.meta = []
self.db_server = db_handler()
self.classify()
self.save_results()
self.print_results()
#self.view_db()
#global db_server
def classify(self):
counter = 0
for videoPath in self.media:
startTime = int(round(time.time()))
dim = (192, 144)
resize_flag = False
vid = videoPath.split("/")[-1]
print("### {}/{} Processing {} ###".format(counter, len(self.media), vid))
video = cv2.VideoCapture(videoPath)
#fps = video.get(cv2.CAP_PROP_FPS)
#res = (video.get(cv2.CAP_PROP_FRAME_WIDTH),video.get(cv2.CAP_PROP_FRAME_HEIGHT))
width, height, duration, fps = findVideoMetada(videoPath)
res = (width, height)
self.meta.append({"res" : res, "fps" : fps, "duration" : duration})
if res[1] > 144:
r = res[0]/res[1]
dim = (int(144 * r), 144)
resize_flag = True
counter += 1
suc, img = video.read()
count_1 = 0
count_2 = 1
img_processed = []
while suc:
if count_1%int(fps) == 0:
img_g = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
if resize_flag:
img_g = cv2.resize(img_g, dim)
img_processed.append(img_g)
count_2 += 1
count_1 += 1
suc, img = video.read()
#print("\tFPS: {} # Resolution: {} #: New Resolution: {} # Numer of frames: {}".format(round(fps,2), res, dim, len(img_processed)))
x = 3 # Number of following seconds to compare to
ratio = []
points = [i for i in range(0,len(img_processed), int(len(img_processed)/4))[:4]]
points.append(len(img_processed))
def procces (start,end,count):
for i in range(start, end):
img1 = img_processed[i]
for j in range(count):
i_2 = i + j
if i_2 < len(img_processed):
img2 = img_processed[i+j]
tmp = self.compare_images(img1, img2)
ratio.append(tmp)
threads = []
threads.append(Thread(target=procces, args=(points[0], points[1], x)))
threads.append(Thread(target=procces, args=(points[1], points[2], x)))
threads.append(Thread(target=procces, args=(points[2], points[3], x)))
threads.append(Thread(target=procces, args=(points[3], points[4], x)))
for t in threads:
t.start()
for t in threads:
t.join()
avg = sum(ratio) / float(len(ratio))
avg = round(avg,2)
if avg >= 0.76:
self.results.append(("S",avg))
else:
self.results.append(("N",avg))
endTime = int(round(time.time()))
self.times.append(endTime-startTime)
def save_results(self):
for i, videoPath in enumerate(self.media):
vid = videoPath.split("/")[-1]
data = {'name' : vid, 'path' : videoPath, 'kullanici': 'kubra', 'class': self.results[i][0], 'meta' : self.meta[i]}
self.db_server.save(db_vc, data, doc_id=videoPath)
def print_results(self):
print("\n\n### Results ###\n")
correct_counter = 0
for i, videoPath in enumerate(self.media):
vid = videoPath.split("/")[-1]
if self.tags:
if self.results[i][0] == self.tags[vid]:
correct_counter += 1
print("# {} # Result: {} ({}) # Actual: {} # Time: {}".format(vid, self.results[i][0], self.results[i][1], self.tags[vid], self.times[i]))
print("correct answer rate: {}/{}".format(correct_counter, len(self.media)))
else:
print("# {} # Result: {} ({}) # Time: {}".format(vid, self.results[i][0], self.results[i][1], self.times[i]))
def mse(self, imageA, imageB):
err = np.sum((imageA.astype("float") - imageB.astype("float")) ** 2)
err /= float(imageA.shape[0] * imageA.shape[1])
return err
def compare_images(self, imageA, imageB, algo="ssim"):
if algo == "ssim":
m = ssim(imageA, imageB) # Better results -- Structural Similarity
else:
m = self.mse(imageA, imageB) # Faster results -- Mean squared error
return m
# function to find the resolution of the input video file
def findVideoMetada(pathToInputVideo):
cmd = "ffprobe -v quiet -print_format json -show_streams"
args = shlex.split(cmd)
args.append(pathToInputVideo)
# run the ffprobe process, decode stdout into utf-8 & convert to JSON
ffprobeOutput = subprocess.check_output(args).decode('utf-8')
ffprobeOutput = json.loads(ffprobeOutput)
# prints all the metadata available:
#import pprint
#pp = pprint.PrettyPrinter(indent=2)
#pp.pprint(ffprobeOutput['streams'][0])
# for example, find height and width
width = int(ffprobeOutput['streams'][0]['width'])
height = int(ffprobeOutput['streams'][0]['height'])
duration_ = float(ffprobeOutput['streams'][0]['duration'])
duration = str(datetime.timedelta(seconds=duration_))
try:
nb_frames = int(ffprobeOutput['streams'][0]['nb_frames'])
fps = nb_frames/duration_
except:
try:
avg_frame_rate = ffprobeOutput['streams'][0]['avg_frame_rate']
tmp = avg_frame_rate.split('/')
except:
r_frame_rate = ffprobeOutput['streams'][0]['r_frame_rate']
tmp = r_frame_rate.split('/')
fps = float(tmp[0])/float(tmp[1])
print(width, height, duration, fps)
return width, height, duration, fps