影片
\(t\)
CNN
CNN
CNN
CNN
圖片特徵
圖片特徵
圖片特徵
圖片特徵
RNN
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
音效特徵
資料庫片段
資料庫片段
資料庫片段
資料庫片段
\rbrace
\rbrace
\rbrace
\rbrace
\rbrace
資料庫片段
下載圖片
圖片前處理
蒐集訓練資料
建構類神經網路
訓練
討論模型表現
訓練模型
擷取畫面
標記畫面
根據標記配音效
影片配音效
調整模型
根據標記配音效
討論成效
成果
1.原始網路
2.去除決策網路
3.訓練決策網路
4.微調
ImageNet
extraction
decision
prediction1
20 classes
extraction
decision
prediction2
transfer
(retrain)
ResNet152V2
(1000 classes)
new model
frames
CNN
fire
labels
ResNet
ALGORITHM
fire
ResNet
-label-
ResNet
-label-
ResNet
ambulance
ResNet
ambulance
ResNet
ambulance
fire
soundeffect
audios
\text{Video}\begin{cases}\begin{array}{llll}
\text{Img}_1 & \xrightarrow{\text{ResNet}} & \text{Class}_1\\
\text{Img}_2 & \xrightarrow{\text{ResNet}} & \text{Class}_2\\
\vdots & & \vdots\\
\text{Img}_{t-1} & \xrightarrow{\text{ResNet}} & \text{Class}_{t-1}\\
\text{Img}_t & \xrightarrow{\text{ResNet}} & \text{Class}_t\\
\end{array}\end{cases}
\text{Video}\begin{cases}\begin{array}{llll}
\text{Img}_1 & \xrightarrow{\text{ResNet}} & \text{Class}_1\\
\text{Img}_2 & \xrightarrow{\text{ResNet}} & \text{Class}_2\\
\vdots & & \vdots\\
\text{Img}_{t-1} & \xrightarrow{\text{ResNet}} & \text{Class}_{t-1}\\
\text{Img}_t & \xrightarrow{\text{ResNet}} & \text{Class}_t\\
\end{array}\end{cases}
\(\text{Class}_t:\)
t=1\qquad A
\vdots
\(\text{Sound}_A\)
\(\text{Sound}_B\)
t=2\qquad A
t=3\qquad A
t=4\qquad A
t=5\qquad B
t=6\qquad B
舊畫面人物
新畫面人物
cost (distance)
vanish cost
消失節點
vanish cost
cost (distance)
舊畫面人物
消失節點
新畫面人物
\(e^-\)
\(e^-\)
\(e^-\)
\(\gamma\)
\(\gamma\)
\(Q_e\)
\(Q_e\)
\(\gamma\)
\(e^+\)
\(e^+\)
\(e^-\)
\(e^-\)
\(Q_e\)
\(Q_e\)
\(W^+\)
\(\nu_e\)
\(e^-\)
\(e^-\)
\(\nu_e\)
\(a\)
\(\sqrt{3}a\)
arr[0]
arr[1]
...
arr[ptr-1]
arr[ptr]
arr[ptr+1]
\(\uparrow\)
ptr
Stack data
arr[head]
arr[head+1]
...
...
arr[tail-1]
arr[tail]
\(\uparrow\)
tail
Queue data
arr[head-1]
arr[tail+1]
\(\uparrow\)
head
prev
next
data
prev
next
data
prev
next
data
prev
next
data
prev
next
data
NULL
NULL
Node
Node
Node
Node
Node
Linked List data
prev
next
data
prev
next
data
prev
next
data
NULL
NULL
Node
Node
Node
Node
Node
prev
next
prev
next
\(\uparrow\)
tail
\(\uparrow\)
head
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
41
34
62
0
45
78
5
58
69
99
0
5
24
58
34
41
78
67
62
69
64
45
0
5
24
58
34
41
78
67
62
69
64
45
17
5
24
58
34
41
78
67
62
69
64
45
before
after
\((t, p)\)
\((t', p')\)
\(T\)
before
after
\((t, p)\)
\((t', p')\)
\(T\)
\(T'\)
before
after
\((t, p)\)
\((t', p')\)
\(T\)
\(T'\)
before
after
\((t', p')\)
\((t, p)\)
\(T\)
before
after
\((t, p)\)
\((t', p')\)
\(T\)
\(\hat{T}\)
\((\hat{t}, \hat{p})\)
1,3
1,2
1,1
2,3
2,2
2,1
3,3
3,2
3,1
4,1
1,2
1,1
2,3
2,2
2,1
3,3
3,2
4,1
1,2
1,1
2,3
2,2
2,1
3,3
3,2
4,1
-
Wifi名稱：CSIE_guest
-
EAP Method：PEAP
-
Phase 2 authentication：MSCHAPV2
-
身分 Identity：ioic2023
-
匿名身分 Anonymous identity：（免填）
-
密碼 Password：CNS5DS2D
\(D\)
\(A'\)
\(C\)
\(D\)
\(B'\)
\(C\)
\(D\)
\(A\)
\(C'\)
\(D\)
\(A'\)
\(B'\)
\(C'\)
\(C\)
\(A\)
A
BC
0
1
00
01
11
10
A
BC
0
1
00
01
11
10
1
1
1
X
X
X
0
0
AB
CD
00
01
11
10
00
01
11
10
AB
CD
00
01
11
10
00
01
11
10
0
0
0
0
0
0
0
X
X
1
1
1
1
1
1
1
AB
CD
00
01
11
10
00
01
11
10
AB
CD
00
01
11
10
00
01
11
10
0
0
0
0
0
0
0
X
X
1
1
1
1
1
1
1
AB
CD
00
01
11
10
00
01
11
10
AB
CD
00
01
11
10
00
01
11
10
1
1
1
0
0
0
X
X
1
0
1
1
1
1
1
0
AB
CD
00
01
11
10
00
01
11
10
AB
CD
00
01
11
10
00
01
11
10
0
0
0
0
0
0
0
1
0
1
1
1
1
0
0
0
AB
CD
00
01
11
10
00
01
11
10
0
1
1
1
1
0
0
1
1
1
1
0
1
0
0
1
AB
CD
00
01
11
10
00
01
11
10
0
1
1
1
1
0
0
1
1
1
1
0
1
0
0
1
S_0
S_1
S_2
S_3
\left\{A, C\right\}
\(C_1\)
\left\{A\right\}
\(C_2\)
\left\{B, C\right\}
\(C_3\)
\left\{B, C\right\}
\(C_4\)
\left\{A, B\right\}
\(C_5\)
\left\{C\right\}
\(C_1\)
\left\{A\right\}
\(C_2\)
\left\{B, C\right\}
\(C_3\)
\left\{B, C\right\}
\(C_4\)
\left\{A, B\right\}
\(C_5\)
Alice
pizza
quesadillas
ramen
sushi
Chris
pizza
quesadillas
ramen
Bob
pizza
ramen
=
David
pizza
ramen
sushi
\neq
\neq
\neq
\neq
17
1
13
6
9
9
\(\alpha, v, \beta\)
\(9, 9, \infty\)
\(-\infty, 9, 9\)
\(17, 17, 9\)
\(9, 9, \infty\)
\(-\infty, 17, 9\)
\(9, 7, 7\)
\(9, 7, \infty\)
\(\alpha, v, \beta\)
\(9, 9, \infty\)
\(-\infty, 6, 6\)
\(6, 6, \infty\)
\(13, 13, 6\)
\(-\infty, 13, 6\)
\(7, 7, 6\)
\(17, 17, \infty\)
\(6, 9, 9\)
\(9, 9, 17\)
\(15, 17, 17\)
\(15, 11, 11\)
\(6, 15, 15\)
Max
Expected
Expected
Expected
Expected
\(\cdots\)
fold
call
raise 2
raise 1024
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
Min
inode
address0
address1
address2
\(\vdots\)
address10
address11
address12
data
data
address0
address1
\(\vdots\)
address4
address255
\(\vdots\)
address0
\(\vdots\)
address94
address255
\(\vdots\)
data
66666 = 11 + 66655 = 11 + 65536 + 1119 = 11 + 65536 + 1024 + 95
address0~255
address0~255
pretrained
transformer
Scalar
Token
Textual
sentence
representation
\(W_{768\times3}\)
ReLU
average
scroing
average
dancibility
fill missings
normalization
Deep Learning Model
Dancibility
Raw Data
Trainables
Model Inputs
normalized
features
系上資訊站
工作坊
讀document會
訪談
新生手冊
課程完成度須達 100%
解題進度須達標準
須蓋章完成
追蹤社團 IG
username
轉發貼文
client
dash.mpd
server
setup
playing
init-stream1.m4s
chunk-stream1-00001.m4s
chunk-stream1-00002.m4s
chunk-stream1-00003.m4s
chunk-stream1-00004.m4s
long seek / back seek / change resolution
0
7
12
15
20
25
32
Hazard
IF
ID
EX
MEM
WB
Fwd
ID
IF
HD
EX
EX
Messenger
Line
Chrome
+
=
Default Weekly Schedule
Original
Result
picture factory
By thomaswang2003
